Vous êtes sur la page 1sur 15

Personal Relationships, 15 (2008), 391405. Printed in the United States of America.

Copyright 2008 IARR. 1350-4126=08

A general model of relationship commitment:


Evidence from same-sex partners

LAWRENCE A. KURDEK
Wright State University

Abstract
A model of relationship commitment previously validated with data from both partners from dating heterosexual
couples was tested with survey data obtained from both partners from 304 same-sex couples cohabiting in the
United States. The model posits that commitment is influenced by factors that are ordered along a proximaldistal
continuum. From most distal to most proximal, these factors are personality traits, support for the relationship from
family members and friends, effective arguing, and dependence on the relationship. Of the 25 predicted effects, 23
were significant (p , .05). Findings support the use of the model for understanding commitment processes for diverse
types of dyadic relationships.

What factors affect the extent to which someone will want to stay in a relationship, even
when the quality of that relationship is poor?
Most scholars agree that multiple factors influence relationship commitment. These factors
include the personality traits of each relationship partner (intrapersonal factors), the extent
to which the relationship is embedded within
other relationships (contextual factors), the
quality of interactions between partners (interpersonal factors), and the extent to which each
partner regards the relationship as meeting
personal needs (interdependence factors;
Adams & Jones, 1997; Huston, 2000; Karney
& Bradbury, 1995; Rusbult, Martz, & Agnew,
1998). To date, the focus of most conceptual
models of commitment (such as the investment model of Rusbult et al., 1998) has been
on one set of these factors. Kurdek (2008)
recently integrated these factors in a single-

Lawrence A. Kurdek, Department of Psychology, Wright


State University.
I thank the couples who participated in this study,
Joseph Olsen for clarifying aspects of data analysis, and
the anonymous reviewers for their helpful comments.
Correspondence should be addressed to Larry Kurdek,
Department of Psychology, Wright State University, Dayton, OH 45435-0001, e-mail: larry.kurdek@wright.edu.

path model and tested predictions derived


from that model with data collected from both
partners from dating heterosexual couples at
one assessment (i.e., the data were not longitudinal). In the current article, I evaluate the
generality of this model by testing its applicability to nonlongitudinal data collected from
both partners from cohabiting gay male and
lesbian couples.
Components of the model
As shown at the top of Figure 1, the model has
five components. The first and the fourth components include more than one variable.
Because the couple is the unit of analysis in
the model, there is a set of paths involving
eight variables for each relationship partner.
The end-point component at the far right of
the figure references relationship commitment, the intent to stay in a relationship. The
other four components represent influences on
commitment that are time ordered along a distalproximal continuum. The theoretical basis
for this ordering is that distal components capture far influences on commitment, such as
those that are brought to the relationship,
whereas proximal components capture near

391

392

L. A. Kurdek
Component 1

Component 2

Component 4

Component 3

Personality traits

Component 5

Dependence on the
relationship

Distal

Proximal

Partner 1
satisfaction

neuroticism

d
f
support

effective
arguing

alternatives

commitment

expressiveness
investment

Partner 2
satisfaction

o'
d'
f'
support

a'

j'

g'

m'

neuroticism

e'

n'

effective
arguing

k'

h'

commitment

alternatives

l'

i'

expressiveness
investment

Figure 1. Predicted links from an integrative model of relationship commitment ordered along
a distalproximal continuum.
influences on commitment, such as those
developed within the relationship itself.
The most distal link to commitment within
the model concerns personality traits, which
are stable predispositions that antedate the
relationship and affect how partners behave
in any relationship (Karney & Bradbury,
1995). The two specific traits of interest were
neuroticismthe tendency to experience negative affect such as anxiety, depression, and
hostility (Costa & McCrae, 1992)and

expressivenessthe disposition to show stereotypically feminine qualities such as kindness, gentleness, and understanding (Miller,
Caughlin, & Huston, 2003). These two traits
are consistent predictors of relationship quality, with high levels of neuroticism linked to
low levels of commitment (Kurdek, 1997) and
high levels of expressiveness linked to high
levels of satisfaction (Miller et al., 2003).
The second and more proximal link to commitment concerns support for the relationship.

Model of relationship commitment

I include this component because intimate


relationships coexist with and are affected by
personal relationships involving family members and friends in particular (Huston, 2000;
Milardo & Helms-Erikson, 2000). Although
the level of support for a specific relationship
likely depends on particular aspects of that
relationship, the personal history that each
relationship partner has with support providers
before the relationship began also affects it
(Sarason, Sarason, & Shearin, 1986). Thus,
some providers might support the relationship
as an extension of the relationship partner (e.g.,
Ill support whatever makes her happy.).
Etcheverry and Agnew (2004) found positive
links between perceived support for ones relationship and commitment to that relationship.
The third and even more proximal link to
commitment concerns effective arguing between partners. Unlike the previous two components, which include influences that exist
before the relationship, this component addresses
only interactions between partners in the specific relationship. Although other kinds of partner interactions occur, Gottman et al. (2003)
regarded the extent to which partners effectively resolve conflict as a key determinant
of change in relationship satisfaction as well
as relationship stability. Rusbult, Verette,
Whitney, Slovik, and Lipkuss (1991) finding
of a relation between the tendency to react
constructively to a partners transgressions
and the commitment to the relationship and
Kurdeks (1994b) finding that decreases in
effective arguing predict relationship dissolution led me to expect a link between frequent
effective arguing and high commitment.
The fourth and most proximal link to commitment concerns dependence on the relationship. This component is the most proximal link
to commitment in that it directly indexes the
extent to which partners rely on the specific
relationship to fulfill important needs. I drew
three specific sources of dependence from
Rusbult and colleagues (1998) well-validated
investment model. These included level of satisfaction with the relationship, the attractiveness of alternatives to the relationship, and the
extent of investments in the relationship.
Meta-analytic evidence has documented that
each of these three sources of dependence

393

accounts for unique variance in commitment


(Le & Agnew, 2003).
The arrows in Figure 1 show the predicted
links between variables within the model.
These links represent three kinds of effects that
include nondirectional links between the personality traits, directional links between adjacent
variables within the model, and directional
links between personality traits and nonadjacent variables. The initial validation of the
model supported each of these links (Kurdek,
2008).
The model involves four assumptions
regarding these links. First, the predicted links
are not necessarily causal in nature. Rather, all
variables within the model predict commitment, although some more distally and some
more proximally than others. Second, the most
direct links are between variables that are adjacent to each other in terms of their timeordered influence on the relationship. Third,
I posited no links between personality traits
and commitment, support and commitment,
and effective arguing and commitment as I
anticipated that relevant intervening variables
along the distalproximal continuum would
mediate these links. That is, I predicted that
effects for relatively distal variables would
be nonsignificant net of the effects associated
with relatively proximal variables. Finally, in
instances where an outcome variable has more
than one predictor, the predicted effects refer
to the unique effect that each predictor makes
with controls for other relevant predictors.
Next, I present the predictions regarding each
link. Unless noted otherwise, supportive evidence comes from studies in which authors
provided no information about participants
sexual orientation.
Reciprocal intrapartner and cross-partner
links regarding the personality traits
The first set of predictions involved nondirectional links for the two personality traits for
each partner. The left side of Figure 1 illustrates these with double-headed curved arrows
(Link a for Partner 1 and Link a9 for Partner 2).
Consistent with evidence that neuroticism
and expressiveness covary within persons
(Costa & McCrae, 1992), I expected an inverse

394

relation between these two personality traits


within each partner. Consistent with evidence
that married heterosexual partners are similar
to each other on personality traits due to assortative mating (Gonzaga, Campos, & Bradbury,
2007), I predicted positive links between partners for both neuroticism (Link b) and expressiveness (Link c).
Directional links between adjacent variables
The second set of predictions referred to directional links between adjacent variables in the
model. As already noted, I anticipated links
between distal variables and the next-most proximal variable(s) in the model. Specific links
between adjacently ordered variables are named
in Figure 1 with letters d through l for Partner
1 and with letters d9 through l9 for Partner 2.
Controlling for expressiveness, I expected
a negative link between neuroticism and support for the relationship (Links d and d9). In
addition, partialing out the effects of neuroticism, I predicted a positive link between
expressiveness and support for the relationship
(Links e and e9). These predictions are consistent with reports of a negative link between
neuroticism and perceived adequacy of support (Bolger & Eckenrode, 1991) and a positive
link between expressiveness and perceived
adequacy of support (Roos & Cohen, 1987).
In turn, I expected a positive link between
support for the relationship and effective arguing (Links f and f9). The rationale for this prediction comes from evidence from gay male,
lesbian, and heterosexual partners that problems with family and friends are frequent areas
of conflict (Kurdek, 1994a; Sanford, 2003) and
that there is a negative link between satisfaction with social support and levels of disagreement (Kurdek, 1992).
In turn, I predicted links between effective
arguing and each of the three markers for
dependence on the relationship. I anticipated
a positive link for satisfaction (Links g and g9),
a negative link for quality of alternatives
(Links h and h9), and a positive link for investment (Links i and i9). Support for these predictions come from findings from gay male,
lesbian, and heterosexual partners of positive
relations between effective arguing and satis-

L. A. Kurdek

faction with the relationship (Kurdek, 1994b)


and findings of positive relations between constructive reactions to a partners transgressions
and dependence on the relationship as indexed
by high satisfaction, low attraction to alternatives to the relationship, and high investment
in the relationship (Rusbult et al., 1991).
Finally, I predicted that each of the three
markers for dependence on the relationship
would account for unique variance in commitment. I based this prediction on meta-analytic
evidence from gay male, lesbian, and heterosexual partners that high satisfaction (Links j
and j9), low quality of alternatives (Links k and
k9), and high investment (Links l and l9)
account for unique portions of variance in high
levels of commitment (Le & Agnew, 2003).
Directional links between personality traits
and nonadjacent variables
In the initial validation of the model (Kurdek,
2008), although I found confirmation of
expected reciprocal links involving the personality traits and directional links between adjacent variables with data from both partners
from dating heterosexual couples, modification indices from structural equation modeling
analyses indicated that adding two sets of links
would improve the fit of the model. The first
set involved directional links between personality traits and nonadjacent variables, and the
second set involved directional cross-partner
links. The links involving personality traits
and nonadjacent variables are part of the
model tested in the current study because these
links are conceptually meaningful and documented by previous work. In contrast, crosspartner links were not part of the model because
few cross-partner links were significant in the
original validation of the model. Nonetheless,
I examined the modification indices associated with the model in Figure 1 to see if crosspartner links warranted inclusion.
The final set of predictions involved directional links between personality traits and two
nonadjacent components of the modeleffective arguing and satisfaction with the relationship. These predictions indicate that support
did not mediate the links between personality
traits and effective arguing and that support or

Model of relationship commitment

effective arguing did not mediate the links


between personality traits and satisfaction.
That is, I anticipated that personality traits
would exert unique effects even partialling
out the influences of relevant intervening variables from the model.
With controls for expressiveness and support, I expected negative links between neuroticism and effective arguing (Links m and m9).
Net of effects for neuroticism and support, I
predicted positive links between expressiveness and effective arguing (Links n and n9).
The basis for these predictions comes from
evidence from gay male, lesbian, and heterosexual partners of links between personality
traits and conflict resolution styles (Bradbury,
Campbell, & Fincham, 1995; Kurdek, 1991,
1997; McGonagle, Kessler, & Schilling,
1992). Neuroticism and expressiveness might
account for unique variance in effective arguing independent of support because these traits
are likely to influence how partners manage
conflict in areas unrelated to the social support
system (Kurdek, 1994a; Sanford, 2003). The
predicted unique importance of neuroticism
and expressiveness for conflict resolution is
consistent with the view that personality traits
are enduring predispositions that act as either
vulnerabilities (neuroticism) or strengths
(expressiveness) in affecting the adaptive processes needed for the development of healthy
relationships (Gonzaga et al., 2007).
Separate from the effects of effective arguing, I anticipated positive links between
expressiveness and satisfaction with the relationship (Links o and o9). I based this prediction on Miller et al. (2003) finding of links
between expressiveness and satisfaction net
of effects for affectionate behavior and responsiveness, which are likely correlates of effective arguing. Although neuroticism is a wellestablished predictor of relationship satisfaction, I predicted a dilution of its unique predictive power with regard to satisfaction when
I also included effects for both expressiveness
and effective arguing.
Predicted links not of central interest
Although not shown in Figure 1 because they
were not of focal interest in the current study,

395

the model also included two sets of correlated


errors that addressed aspects of nonindependence among variables from the model. These
correlated errors occurred for both male and
female partners from dating heterosexual couples in the original validation sample (Kurdek,
2008) and were conceptually and methodologically meaningful (Cole, Ciesla, & Steiger, 2007).
The first set involved correlated errors
between partners from the same couple for
each of the six outcome variables. Allowing
for correlated errors between partners from
the same couple is the standard way to control
for the nonindependence between partners
scores (Kenny, Kashy, & Cook, 2006). The
second set involved the three within-partner
correlations involving satisfaction, alternatives,
and investment and the one within-partner
correlation involving satisfaction and support.
I assumed correlated errors for satisfaction,
alternatives, and investment because these
scores came from the same measure that used
the same response scale. I anticipated correlated errors between satisfaction and support
because each involved a general assessment of
the quality of a relationship, one focusing on
an intimate partner (satisfaction) and the other
on members of ones support system (support).
As such, a common response bias might influence each score (Kline, 2005).
Purpose of the present study
Previous evidence (Kurdek, 2004, 2006) documents that the predictors of satisfaction, commitment, and stability are generally the same
for gay male, lesbian, and heterosexual partners. No researchers, however, have evaluated
the generality of complex path models that
integrate the influences of intrapersonal factors, contextual factors, the quality of interactions between relationship partners, and the
extent to which the relationship meets personal
needs on commitment across different types of
couples. I initially tested the path model proposed here with data from both partners from
dating heterosexual couples because much of
the evidence used to support predictions from
the model comes from heterosexual partners.
If this model is a general model of relationship
commitment that is applicable to diverse types

396

of relationships, then predictions derived from


it also should hold for partners from same-sex
couples. Evidence supporting a general model
is important because it is consistent with the
view that a core set of processes regulates
dyadic relationships, regardless of the type of
intimate partner (Kurdek, 2004).
Method
Participants
I sent surveys to each of the 610 couples (437
lesbian) who had applied for civil union certificates in the state of Vermont in the United
States in 2004. Surveys were not deliverable
for 39 couples (27 lesbian), 52 couples (37
lesbian) declined to participate, 9 couples (3
lesbian) had separated, and partners from 3
couples (2 lesbian) were not living together
full time. From the remaining 139 gay male
and 398 lesbian couples, both partners for 45
gay male and 154 lesbian couples returned
completed surveys, for return rates of 32%
and 39%, respectively. Balsam, Beauchaine,
Rothblum, and Solomon (2008) reported similar rates in that 42% of the 2,269 Vermont
same-sex couples they contacted from certificates issued from July 1, 2000, to June 30,
2001 agreed to participate in their study and
82% of the first 400 couples sampled returned
completed surveys.
Partners who returned surveys nominated
partners from other same-sex couples who
might be interested in participating in the
study. Of the 277 couples (202 lesbian) contacted as a result of these nominations, both
partners from 36 gay male couples and 69 lesbian couples returned completed surveys, for
return rates of 48% and 34%, respectively.
Thus, the participants in the final sample
included both partners from 81 gay male and
223 lesbian couples who did not receive payment for participation. Most of the participants
were White (91.1%), 27.0% had earned a baccalaureate degree, 78.5% were employed full
time, and 18.7% had children living with them
full time. Their modal annual personal income
(for 10.3% of the sample) was between
US$35,000 and US$39,999. Partners mean
age was 41.13 years; they cohabited a mean

L. A. Kurdek

of 8.57 years, and 78.9% of them had obtained


a civil union in the state of Vermont.
Measures of demographic variables
Participants reported their age in years and
described their race as White, Black, Hispanic,
Asian, or other. They indicated their highest
level of formal education completed as one
of seven categories (1 less than eighth
grade, 7 doctoral degree) and reported
their own total annual income as one of 24
categories (1 less than U.S.$5,000, 24
more than U.S.$115,000) They also reported
the years of cohabitation with their partner,
their employment status (unemployed, retired,
part-time employed, or full-time employed),
and whether (1 yes, 2 no) they had
entered into a civil union or had children living
with them on a full-time basis.
Measures of variables from the relationship
commitment model
Personality traits.
I measured neuroticism
with 12 items (e.g., I often feel inferior to
others) from the NEO Five-Factor Inventory
(Costa & McCrae, 1992) using a 5-point
response scale (1 strongly disagree, 5
strongly agree) I measured expressiveness
with 9 items (e.g., gentle) derived from
Kurdeks (1987) principal components analysis of Bems (1974) Sex Role Inventory. Participants used a 7-point response scale to
indicate how true each characteristic was of
them (1 never or almost never true, 7
almost or almost always true) I used Cronbachs alpha (a) to assess the reliability of all
averaged composite scores. Cronbachs alphas
were .84 for neuroticism and .91 for
expressiveness.
Support for the relationship.
Participants
used a 9-point scale (1 not at all, 9 quite
a lot) to rate perceptions of support for their
relationship from each of five sources: members of their own family, members of their
partners family, their own friends, their partners friends, and friends in common with their
partner (e.g., To what degree does your family approve and support your relationship?).

Model of relationship commitment

The first four items came from Sprecher and


Felmlees (1992) measure of social reactions,
and I added the last item (a .74).
Effective arguing.
I assessed effective
arguing with eight items (e.g., Overall, our
arguments are brief and quickly forgotten)
from Kurdeks (1994b) Ineffective Arguing
Inventory. Participants used a 5-point format
(1 disagree strongly, 5 agree strongly;
a .89).
Dependence on the relationship.
Participants used a 9-point response format (0 do
not agree at all, 8 agree completely) to
indicate how much they agreed with items
from Rusbult and colleagues (1998) Investment Model Scale. There were four items
apiece for satisfaction (e.g., I feel satisfied
with my relationship), quality of alternatives
(e.g., The options to this relationship [dating
someone else, spending time alone or with
friends, etc.] are attractive), and investment
size (e.g., I feel involved in this relationship,
and feel that I have put a great deal into it; as
.90, .84, and .68, respectively).
Commitment.
I measured commitment with
eight items (e.g., I am committed to maintaining my relationship with my partner) from
Sternbergs (1988) measure of commitment. Participants used a 9-point response format (1 not
at all true, 9 extremely true; a .89).
Results
Bivariate links
For descriptive purposes, I present intraclass
correlations involving both own and partner
variables in Table 1 for the total sample. I conducted significance tests with procedures
described by Griffin and Gonzalez (1995).
Of particular note is that the correlation
between commitment and each of its seven
predictors was significant and in the expected
direction and that the predictors themselves
were interrelated. In order to see if I could
ignore type of same-sex couple as a variable
that moderated the predicted links, I compared
the intraclass variancecovariance matrix

397

based on the 16 variables shown in Table 1


for gay male partners to that for lesbian partners. A two-group structural equation modeling analysis indicated that the elements of
the two matrices did not differ, v2 (151, N
304) 168.90, p . .05. Consequently, I combined partners from gay male couples and partners from lesbian couples into one sample. The
separate matrices for gay male partners and for
lesbian partners are available upon request.
Plan of analysis
I tested the predicted model shown in Figure 1
with structural equation modeling procedures
Olsen and Kenny (2006) developed for data
from dyads in which members are not distinguished from each other. The analysis of data
from both members from the same couple
requires special procedures, which take the
nonindependence of partners scores into
account (Kenny et al., 2006). In the current
study, the interpartner correlations (see own
partner links in Table 1) for each of the eight
variables in the model were positive and significant, indicating that partners scores were
not independent of each other.
Olsen and Kennys (2006) procedures first
involved setting up the data for Partner 1 and
Partner 2 as if they were distinguishable members (Figure 1). Then, I constrained six types
of parameters in the model to be equal for each
partner. These parameters included regression
coefficients, means for predictors, variances
for predictors, intercepts for outcomes, residual variances for outcomes, and within-person
correlated errors. Because I constrained these
parameters to be equal for both partners, I
needed only one estimate for each parameter.
For example, with regard to the regression
coefficients relevant to the links shown in Figure 1, Path a was set to be equal to Path
a9 because there was no reason to expect that
these coefficients would differ for partners
from same-sex couples assigned randomly to
Partner 1 or Partner 2 status. Testing the model
with partners who were not distinguished from
each other also involved estimating means and
variances for neuroticism and expressiveness
(which only served as predictors) and estimating intercepts and residual variances for

10

11

12

13

14

15

2.17*
2.17*
.16*
.30*
.25*
.28*
2.24*
.27*
.23*
.53*
.05 2.09* 2.02 2.20* 2.31*
.01
.11*
.16*
.11*
.28* 2.23*
2.17*
.25*
.25*
.47*
.63* 2.36*
.28*
.19* 2.06 2.19* 2.20* 2.14*
.06 2.02 2.08*
2.06
.11*
.10*
.16*
.13* 2.06
.08*
.08* 2.17*
2.19*
.10*
.64*
.18*
.11* 2.03
.16*
.14* 2.17*
.16*
2.20*
.16*
.18*
.59*
.32* 2.14*
.09*
.26* 2.30*
.25*
.28*
2.14*
.13*
.11*
.32*
.39* 2.21*
.13*
.26* 2.23*
.28*
.23* 2.53*
.06 2.06 2.03 2.14* 2.21*
.32* 2.14* 2.18*
.05 2.09* 2.02
.20* 2.31*
2.03
.08
.16*
.09*
.13* 2.14*
.35*
.14*
.01
.11*
.16* 2.11*
.28* 2.23*
2.09*
.08
.14*
.26*
.26* 2.18*
.14*
.26* 2.17*
.25*
.25* 2.47*
.64* 2.36* .28*

*p , .05, according to tests described by Griffin and Gonzalez (1995).

1. O Neuroticism
2. O Expressiveness
3. O Support
4. O Effective arguing
5. O Satisfaction
6. O Alternatives
7. O Investment
8. O Commitment
9. P Neuroticism
10. P Expressiveness
11. P Support
12. P Effective arguing
13. P Satisfaction
14. P Alternatives
15. P Investment
16. P Commitment

Variable

Table 1. Intraclass correlations among own (O) and partner (P) variables for the total sample (N 304)

398
L. A. Kurdek

Model of relationship commitment

support, effective arguing, satisfaction, alternatives, investment, and commitment (which


served as outcome scores). Although these
estimates were part of the model, they were
not of substantive interest in this study.
Fit indices for structural equation modeling
with interchangeable dyads currently include
the model chi-square test, incremental fit indices (such as the comparative fit index [CFI]),
which assess the relative improvement in fit of
the model compared to a null model which
assumes covariances of 0 among all observed
variables, and the root mean square error of
approximation (RMSEA), which is a parsimony-adjusted index with a built-in correction
for model complexity. Because analyses based
on large samples typically yield significant
model chi-square values, the latter two fit indices were of greater interest. Based on Klines
(2005) recommendations, the criteria for good
model fit include a CFI value of .90 or better
and an RMSEA value between .05 and .08.
Olsen and Kenny (2006) provide formulas
for obtaining these fit indices.
Predicted effects from the model
The model provided an acceptable fit to the
data, v2(39, N 304) 126.46, p , .01;
CFI 0.95; RMSEA 0.08. I present standardized estimates for each of the four types of
predicted effects in Table 2. As shown in this
table, 23 of the 25 predicted effects were
significant.
Reciprocal, nondirectional links.
Two of the
three predicted effects occurred. Within each
partner, negative links emerged between neuroticism and expressiveness (Link a). Between
partners, positive links emerged for neuroticism (Link b) but not for expressiveness (Link c).
Directional links between adjacent variables
in the model.
Findings for all directional
effects involved an examination of the partial
regression coefficients associated with each
outcome score. In contrast to the correlations
in Table 1, these coefficients represent the
unique influence of each predictor, partialling
out the effects of other relevant predictors.
Eight of the nine predicted effects occurred.

399

Regarding the only unsupported prediction,


the negative link between neuroticism and
support (see Table 1) became nonsignificant
with controls for expressiveness (Link d).
Independent of neuroticism, a positive link
emerged between expressiveness and support
(Link e). Net of effects for both neuroticism
and expressiveness, a positive link emerged
between support and effective arguing (Link
f). With the effects of expressiveness removed,
a positive link emerged between effective
arguing and satisfaction (Link g). A negative
link emerged between effective arguing and
the quality of alternatives to the relationship (Link h), and a positive link emerged between effective arguing and investment in the
relationship (Link i). Finally, with regard to
commitment, a positive link emerged for satisfaction (Link j), a negative link emerged for
quality of alternatives (Link k), and a positive
link emerged for investment (Link l).
Directional links involving nonadjacent variables in the model.
All three predicted effects
occurred. With controls for expressiveness and
support, a negative link emerged between neuroticism and effective arguing (Link m). Net of
effects for neuroticism and support, a positive
link emerged between expressiveness and effective arguing (Link n). Finally, independent of
effective arguing, a positive link emerged between expressiveness and satisfaction (Link o).
Correlated errors.
The six correlated
errors expected between partners and the four
correlated errors expected within partners
occurred. The errors for each of the six variables that functioned as outcome scores in the
model (support, effective arguing, satisfaction,
alternatives, investment, and commitment)
showed interpartner links. Within partners,
correlated errors emerged for the three interdependence scores (satisfaction with alternatives, satisfaction with investment, and
alternatives with investment) as well as
between support and satisfaction.
Modification indices
None of the modification indices associated
with the model indicated that adding regression

400

L. A. Kurdek

Table 2. Standardized coefficients for predicted model parameters


Parameter

Coefficient

Reciprocal links (with link letter from Figure 1)


a. Neuroticism 4 Expressiveness
b. NeuroticismPartner 1 4 NeuroticismPartner 2
c. ExpressivenessPartner 1 4 ExpressivenessPartner 2
Directional links between adjacent variables (with link letter from Figure 1)
d. Neuroticism / support
e. Expressiveness / support
f. Support / effective arguing
g. Effective arguing / satisfaction
h. Effective arguing / alternatives
i. Effective arguing / investment
j. Satisfaction / commitment
k. Alternatives / commitment
l. Investment / commitment
Directional links between nonadjacent variables (with link letter from Figure 1)
m. Neuroticism / effective arguing
n. Expressiveness / effective arguing
o. Expressiveness / satisfaction
Correlated errors
Between partners
SupportPartner 1 4 SupportPartner 2
Effective arguingPartner 1 4 Effective arguingPartner 2
SatisfactionPartner 1 4 SatisfactionPartner 2
AlternativesPartner 1 4 AlternativesPartner 2
InvestmentPartner 1 4 InvestmentPartner 2
CommitmentPartner 1 4 CommitmentPartner 2
Within partners
Satisfaction 4 Alternatives
Satisfaction 4 Investment
Alternatives 4 Investment
Support 4 Satisfaction

2.16**
.17**
.09
2.04
.11**
.24**
.47**
2.19**
.10**
.56**
2.17**
.08**
2.18**
.12**
.14**

.63**
.55**
.22**
.23**
.29**
.14**
2.18**
.22**
2.17**
.12**

**p , .01.

coefficients would improve the fit of the


model. Thus, I did not add cross-partner links
to the model. This evidence also supports the
models assumption of nonsignificant direct
effects going from personality traits to commitment, from support to commitment, and
from effective arguing to commitment. That
is, relevant intervening variables along the distalproximal continuum mediated the link
between each of these variables and commitment (Table 1). I obtained statistical evidence
regarding mediation by testing the indirect
effects associated with these variables using

the bootstrap method to derive the standard


error for each indirect effect (Shrout & Bolger,
2002). As expected, indirect effects involving
commitment were significant for neuroticism,
expressiveness, support, and effective arguing,
standardized estimates 20.58, 0.12, 0.07,
and .31, respectively, all ps , .01.
Discussion
Although there is a growing literature on the
nature and predictors of relationship quality
for members of gay male and lesbian couples

Model of relationship commitment

(Peplau & Fingerhut, 2007), studies using


these couples to test multivariate models of
relationship quality are scarce. In the current
study, I presented a model that posits that the
influences of personality traits, support for the
relationship from family members and friends,
effective arguing, and dependence on the relationship on commitment can be ordered along
a distalproximal continuum. I initially validated this model with data from both partners
from dating heterosexual couples (Kurdek,
2008). This model extends the scope of previous models such as Rusbult and colleagues
(1998) investment model by including variables with relatively distal influences on commitment. Consistent with the view that general
processes regulate the functioning of dyads
involving differing kinds of intimate partners
(Kurdek, 2004), in the current study, this
model also provided an acceptable level of
fit to data obtained from both partners from
cohabiting same-sex couples. The specific predictions derived from the model reflect three
general kinds of links relevant to commitment,
and I discuss each in turn.
Reciprocal links regarding the personality
traits
The first set of links concerns nondirectional
links involving the personality traits. Consistent with evidence that neuroticism and
expressiveness covary within persons (Costa
& McCrae, 1992), these two personality traits
covaried negatively. Based on evidence from
married heterosexual couples (Gonzaga et al.,
2007), I expected positive links between partners scores for both neuroticism and expressiveness. Only the link for neuroticism
occurred. Using a sample of gay male and lesbian partners different from that used in the
present study, Kurdek (2003) reported significant partner similarity for neuroticism but not
for expressiveness as assessed by measures of
femininity and agreeableness. Because partners from both gay male and lesbian couples
typically report fairly high levels of expressiveness (Kurdek, 1987), the lack of partner
similarity on this variable may be due to
a restricted range of relatively low values on
this variable. Indeed, the mean expressiveness

401

score for the current sample was 5.72, with


a score of 7 being the highest possible score.
Directional links between adjacent variables
in the model
The second set of effects relevant to commitment concerned directional links between
adjacent variables in the model. Eight of the
nine predicted links between distal variables
and the next-most proximal variable in the
model occurred. Previous findings (Roos &
Cohen, 1987) indicate positive links between
expressiveness and the adequacy of overall
levels of perceived social support. In the current study, a positive link emerged between
expressiveness and levels of a more refined
kind of support that specifically targeted support for the relationship by family members
and friends. This link persisted even with controls for neuroticism. Consistent with Bolger
and Eckenrodes (1991) findings, a negative
link emerged between neuroticism and perceived support from family members and
friends. Neuroticism, however, lost its predictive power when the effects for expressiveness
were removed. Thus, of the two expected links
regarding personality traits, the one involving
expressiveness was more robust. Future studies regarding the influence of personality traits
on support for the relationship would do well
to include measures of both neuroticism and
expressiveness so that researchers could assess
their unique effects as well as examine
whether the robust effects of expressiveness
are due to a propensity to engage in affectionate behaviors or a tendency to regard relationship partners more positively than might be
warranted (Miller et al., 2003).
In turn, a positive link occurred between
perceived support for the relationship and
effective arguing net of effects for neuroticism
and expressiveness. This finding underscores
the value of placing intimate relationships
within the context of other social relationships
(Huston, 2000) and accords well with previous
evidence of negative links between the frequency of conflict regarding family and
friends and relationship quality (Kurdek,
1994a; Sanford, 2003) and negative links
between satisfaction with social support and

402

the frequency of disagreements (Kurdek,


1992). It is possible that the unique positive
link between support and effective arguing is
due to the common influence of well-developed
interpersonal skills on both support and effective arguing. That is, partners who enjoy both
support for their relationship from family
members and friends and the effective resolution of conflict in their relationship may have
the personal dispositions and skills that nurture
and maintain interpersonal relationships in
a variety of settings (Rusbult et al., 1991).
In turn, positive links occurred between
effective arguing and each of three markers
of dependence on the relationshiphigh satisfaction (with controls for expressiveness),
low quality of alternatives, and high investment. These results provide further evidence
that partners positive interactions in the area
of conflict resolution bolster each partners
overall dependence on the relationship to meet
important needs (Kurdek, 1994b). In particular, positive interactions between partners in
the area of relationship conflict likely decrease
perceived costs to being in the relationship,
increase the likelihood that partners stay in
the relationship as a result of attractions to
the relationship rather than to an avoidance
of the negative consequences associated with
ending the relationship, and motivate partners
to inhibit destructive reactions to perceived
transgressions (Kurdek, 2007; Rusbult et al.,
1991).
Finally, in agreement with findings from
meta-analyses supporting Rusbult and colleagues (1998) investment model (Le &
Agnew, 2003), each of the three sources of
dependence on the relationship accounted for
unique variance in commitment. Le and
Agnew (2003) reported that across 32 studies,
satisfaction, alternatives, and investment collectively accounted for 61% of the variance in
commitment. In the current study, that value
was 43% but still indicates that the set of interdependence scores accounted for a large portion of the variance in commitment. Although
bivariate links between neuroticism and commitment, expressiveness and commitment,
support and commitment, and effective arguing and commitment were significant, the indirect paths to commitment from these

L. A. Kurdek

variablesbut not the direct effectswere


significant. That is, the bivariate effects were
diminished with controls for the set of interdependence scores in particular. This finding is
consistent with the assumption of the model
that the interdependence scores are the most
proximal predictors of commitment and that
the joint influences of personality traits, support, and effective arguing on commitment are
indirect rather than direct.
Directional links between personality traits
and nonadjacent variables in the model
The third set of links related to commitment
involved the only evidence in this study of
direct paths between a distal component of
the model (personality traits) and nonadjacent
proximal components (effective arguing and
satisfaction). These direct paths support the
view that personality traits are enduring predispositions that act as either vulnerabilities
(neuroticism) or strengths (expressiveness) in
affecting the adaptive processes relevant
to the development of healthy relationships
(Gonzaga et al., 2007).
Both of the predicted links involving effective arguing emerged in that a negative link
occurred between neuroticism and effective
arguing independent of effects for both expressiveness and support, and a positive link
occurred between expressiveness and effective
arguing separate from effects for both neuroticism and support. These findings accord well
with evidence that whereas expressiveness
motivates partners to strive for and to achieve
positive relational goals (Miller et al., 2003),
neuroticism has the opposite effect (Pasch,
Bradbury, & Davila, 1997).
Finally, as expected, a positive link
emerged between expressiveness and satisfaction with the relationship even with controls
for effective arguing. Miller et al. (2003) also
found that expressiveness had direct effects on
satisfaction (independent of affectionate
behavior and responsiveness) and speculated
that expressiveness promotes satisfaction with
the relationship by facilitating partners tendency to idealize their partners. Future studies
could directly test this proposed mediating
effect.

Model of relationship commitment

Regarding links not found


Although having data from both partners from
the same couple enables an examination of
cross-partner effects (Kenny et al., 2006), such
effects did not occur in this study. This null
finding is not too surprising because in the
original validation study involving both partners from dating heterosexual couples, only 2
of the 24 possible cross-partner effects
emerged. It is possible that cross-partner
effects were not salient because the personality
variables are the only truly individual-level
components in the model. Because the other
components of the model involve both partners appraising aspects of the same relationship, intrapartner and cross-partner links are
somewhat redundant. It is also possible that
aspects of ones partner influence decisions
to remain in a relationship, but those aspects
translate into intrapartner processes (such as
level of satisfaction with the relationship) that
primarily drive level of commitment. Thus,
intrapartner links may wash out any unique
effects associated with cross-partner links.
Although these effects are in need of replication, the tentative conclusion is that, at least
within the context of the model presented here,
intrapartner processes regulate commitment
more than cross-partner ones do. Future work
using different measures and methods that do
not rely exclusively on self-report might reexamine whether intrapartner links on commitment are more robust than cross-partner links
on commitment and explore the balance of
distal and proximal links for intrapartner and
cross-partner effects on relationship stability.
Limitations, strengths, and conclusion
This study has several limitations. First, the
sample of gay male and lesbian couples
recruited for this study, many of whom
obtained civil unions in Vermont, was not
likely a representative one. Second, although
some couples had children living with them,
no analyses examined the effects of children
on links from the model. Third, I collected
all data with self-report measures, and I made
no direct observations of effective arguing.
Fourth, because the data were correlational
in nature, results could only highlight the

403

plausibility of the directional influences implied within the model. Fifth, although the
models emphasis on proximal and distal influences is time ordered, I did not have longitudinal data and other types of orderings might
be defended. Sixth, although most of the findings from this study replicate those obtained
with partners from dating relationships, I did
not test the generality of the model statistically
within a structural equation modeling framework by assessing whether model parameters
were equivalent across independent groups of
couples. Finally, although the findings from
this study provide additional evidence that
similar processes regulate the relationships of
same-sex as well as heterosexual partners, I
did not consider predictors of relevance only
to gay male and lesbian partners (such as internalized homophobia and perceived discrimination as a result of sexual orientation; Otis,
Rostosky, Riggle, & Hamrin, 2006) to see if
they added information above and beyond
existing variables in the model. This approach
of adding specialized variables to the core
components of the general model might be
a useful strategy for future researchers wanting
to test the view that variables specific to a particular type of couple really add something
above and beyond variables that appear to
have a general effect on all types of couples.
This study also has some strengths. First,
although there is a sizable literature on the
predictors of commitment, hypotheses based
on systematic and integrative theoretical models are rare. The current model build upon
Rusbult and colleagues (1998) investment
model by considering what relatively distal
factors might be involved in influencing commitment beyond the three most proximate factors derived from the investment model
(satisfaction, quality of alternatives, and
investment). Second, participants were not
heterosexual college students, but rather were
both partners from gay male and lesbian
cohabiting couples, many of whom had
obtained civil unions in Vermont. Finally,
analyses were appropriate for dyad-level data.
Although there was little evidence of crosspartner links, I conducted analyses to see if
such links were important. In conclusion, the
findings from this study provide support for

404

a general model of relationship commitment,


which integrates the way that personality
traits, perceived support for the relationship,
effective arguing, and dependence on the relationship influence commitment to that relationship. The general nature of the model
suggests that it might serve as the basis for
clinical intervention design to repair or to
maintain the relationships of partners from
diverse types of couples.
References
Adams, J. M., & Jones, W. H. (1997). The conceptualization of marital commitment: An integrative analysis.
Journal of Personality and Social Psychology, 72,
11771196.
Balsam, K., Beauchaine, T. P., Rothblum, E. D., &
Solomon, S. E. (2008). Three-year follow-up of samesex couples who had civil unions in Vermont, samesex couples not in civil unions, and heterosexual
married couples. Developmental Psychology, 44, 102116.
Bem, S. L. (1974). The measurement of psychological
androgyny. Journal of Consulting and Clinical Psychology, 45, 155162.
Bolger, N., & Eckenrode, J. (1991). Social relationships,
personality, and anxiety during a major stressful event.
Journal of Personality and Social Psychology, 61,
440449.
Bradbury, T. N., Campbell, S. M., & Fincham, F. D.
(1995). Longitudinal and behavioral analysis of masculinity and femininity in marriage. Journal of Personality and Social Psychology, 68, 328341.
Cole, D. A., Ciesla, J. A., & Steiger, J. H. (2007). The
insidious effects of failing to include design-driven
correlated residuals in latent-variable covariance structure analysis. Psychological Methods, 12, 381398.
Costa, P. T., & McCrae, R. R. (1992). Revised NEO personality inventory (NEO PI-R) and NEO five-factor
inventory (NEO-FFI). Odessa, FL: Psychological
Assessment Resources.
Etcheverry, P. E., & Agnew, C. R. (2004). Subjective
norms and the prediction of romantic state and fate.
Personal Relationships, 11, 409428.
Gonzaga, G. C., Campos, B., & Bradbury, T. N. (2007).
Similarity, convergence, and relationship satisfaction
in dating and married couples. Journal of Personality
and Social Psychology, 93, 3448.
Gottman, J. M., Levenson, R. W., Swanson, C., Swanson,
K., Tyson, R., & Yoshimoto, D. (2003). Observing
gay, lesbian, and heterosexual couples relationships:
Mathematical modeling of conflict interaction. Journal of Homosexuality, 45, 6591.
Griffin, D., & Gonzalez, R. (1995). Correlational analysis
of dyad-level data in the exchangeable case. Psychological Bulletin, 118, 430439.
Huston, T. L. (2000). The social ecology of marriage and
other intimate unions. Journal of Marriage and the
Family, 62, 298320.
Karney, B. R., & Bradbury, T. N. (1995). The longitudinal
course of marital quality and stability: A review of theory,
method, and research. Psychological Bulletin, 118, 334.
Kenny, D. A., Kashy, D. A., & Cook, W. (2006). Dyadic
data analysis. New York: Guilford.

L. A. Kurdek
Kline, R. B. (2005). Principles and practice of structural
equation modeling. New York: Guilford.
Kurdek, L. A. (1987). Sex role self schema and psychological adjustment in coupled homosexual and heterosexual men and women. Sex Roles, 17, 549562.
Kurdek, L. A. (1991). Correlates of relationship satisfaction in cohabiting gay and lesbian couples: Integration
of contextual, investment, and problem-solving models. Journal of Personality and Social Psychology, 61,
910922.
Kurdek, L. A. (1992). Dimensionality of the Dyadic
Adjustment Scale: Evidence from heterosexual and
homosexual couples. Journal of Family Psychology,
6, 2235.
Kurdek, L. A. (1994a). Areas of conflict for gay, lesbian,
and heterosexual couples: What couples argue about
influences relationship satisfaction. Journal of Marriage and the Family, 56, 923934.
Kurdek, L. A. (1994b). Conflict resolution styles in gay,
lesbian, and heterosexual nonparent, and heterosexual
parent couples. Journal of Marriage and the Family,
56, 705722.
Kurdek, L. A. (1997). Relation between neuroticism and
dimensions of relationship commitment: Evidence
from gay, lesbian, and heterosexual couples. Journal
of Family Psychology, 11, 503514.
Kurdek, L. A. (2003). Differences between gay and lesbian
cohabiting couples. Journal of Social and Personal
Relationships, 20, 411436.
Kurdek, L. A. (2004). Are gay and lesbian cohabiting
couples really different from heterosexual married
couples? Journal of Marriage and the Family, 66,
880900.
Kurdek, L. A. (2006). Differences between partners from
heterosexual, gay, and lesbian couples. Journal of
Marriage and Family, 68, 509528.
Kurdek, L. A. (2007). Avoidance motivation and relationship commitment in heterosexual, gay male,
and lesbian partners. Personal Relationships, 14, 291
306.
Kurdek, L. A. (2008). Differences between partners from
Black and White heterosexual dating couples in a path
model of relationship commitment. Journal of Social
and Personal Relationship, 25, 5170.
Le, B., & Agnew, C. R. (2003). Commitment and its theorized determinants: A meta-analysis of the Investment Model. Personal Relationships, 10, 3757.
McGonagle, K. A., Kessler, R. C., & Schilling, E. A.
(1992). The frequency and determinants of marital disagreements in a community sample. Journal of Social
and Personal Relationships, 9, 507524.
Milardo, R. M., & Helms-Erikson, H. (2000). Network
overlap and third-party influence in close relationships. In S. S. Hendrick & C. Hendrick (Eds.), Close
relationships: A sourcebook (pp. 3345). Thousand
Oaks, CA: Sage.
Miller, P. J. E., Caughlin, J. P., & Huston, T. L. (2003).
Trait expressiveness and marital satisfaction: The role
of idealization processes. Journal of Marriage and
Family, 65, 978995.
Olsen, J. A., & Kenny, D. A. (2006). Structural equation
modeling with interchangeable dyads. Psychological
Methods, 11, 127141.
Otis, M. D., Rostosky, S. S., Riggle, E. D. B., & Hamrin,
R. (2006). Stress and relationship quality in same-sex
couples. Journal of Social and Personal Relationships,
23, 8199.

Model of relationship commitment


Pasch, L. A., Bradbury, T. N., & Davila, J. (1997). Gender,
negative affectivity, and observed social support
behavior in marital interaction. Personal Relationships, 4, 361378.
Peplau, L. A., & Fingerhut, A. (2007). The close relationships of lesbians and gay men. Annual Review of Psychology, 58, 405424.
Roos, P. E., & Cohen, L. H. (1987). Sex roles and social
support as moderators of life stress adjustment. Journal
of Personality and Social Psychology, 52, 576585.
Rusbult, C. E., Martz, J. M., & Agnew, C. R. (1998). The
Investment Model Scale: Measuring commitment
level, satisfaction level, quality of alternatives, and
investment size. Personal Relationships, 5, 357392.
Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F., &
Lipkus, I. (1991). Accommodation processes in close
relationships: Theory and preliminary empirical evidence. Journal of Personality and Social Psychology,
60, 5378.

405
Sanford, K. (2003). Problem-solving conversations in
marriage: Does it matter what topics couples discuss?
Personal Relationships, 10, 97112.
Sarason, I. G., Sarason, B. R., & Shearin, E. N. (1986).
Social support as an individual difference variable: Its
stability, origins, and relational aspects. Journal of
Personality and Social Psychology, 50, 845855.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: New procedures
and recommendations. Psychological Methods, 7,
422445.
Sprecher, S., & Felmlee, D. (1992). The influence of
parents and friends on the quality and stability of
romantic relationships: A three-wave longitudinal
investigation. Journal of Marriage and the Family,
54, 888900.
Sternberg, R. J. (1988). Triangulating love. In R. J. Sternberg & M. L. Barnes (Eds.), The psychology of love
(pp. 119138). New Haven, CT: Yale University Press.

Vous aimerez peut-être aussi