INTERPERSONAL RELATIONS AND GROUP PROCESSES

Approach Versus Avoidance: Different Types of Commitment

in Intimate Relationships

Elisabeth Frank and Veronika Brandstätter

University of Munich

The major objective of the present study was to examine whether approach versus avoidance commit-
ment to one’s intimate relationship was differentially predictive of relationship quality parameters in the
long run. In the 1st testing period, 134 participants (67 romantic couples) answered questions about
approach- versus avoidance-related measures. Commitment and relationship quality parameters such as
satisfaction and emotions depending on the partner’s presence were assessed in all 3 testing periods. The
proposed distinction between an approach and an avoidance type of commitment was validated through
correlations with other approach- versus avoidance-related measures. Longitudinal analyses revealed that
approach commitment predicted relationship quality parameters positively, whereas avoidance commit-
ment predicted them negatively. The results are discussed in terms of the benefit of an approach–
avoidance– based conceptualization of commitment.

Commitment to one’s romantic relationship is a key construct in Rusbult, 1991). In the investment model, commitment is concep-
explaining relationship functioning (e.g., M. Johnson, 1991; Lund, tualized as an additive function of (a) satisfaction with the rela-
1985; Lydon, 1996; Rusbult, 1980; Sternberg, 1986). For many tionship, (b) quality of alternatives, and (c) investment size. More
years, researchers in the field of close relationships have been specifically, commitment should increase as the individual feels
defining commitment in many different ways (for an overview, see increasingly satisfied with the relationship, as alternatives decrease
Lydon, 1996). Nevertheless, agreement does at least exist with in quality, and as the magnitude of the individual’s investments in
respect to the view that commitment refers to a specific psycho- the association becomes greater (Rusbult, 1991).
logical state “in which a person feels tied or connected to some- The primary focus of Rusbult’s (1983; Rusbult & Martz, 1995)
one” (Lydon, 1996, p. 192) and that “directly influences P’s [a research was on predicting outcome variables related to the sta-
person’s] decision to continue or end a relationship” (Rusbult, bility of a relationship by directly analyzing the effects of com-
1991, p. 156). Research on commitment in the realm of intimate mitment on the duration of the relationship or on behavior aimed
relationships has mostly been guided by interdependence theory at maintaining the relationship. It has been shown, for instance,
and its extension in Caryl Rusbult’s investment model (e.g., Bui, that the more strongly committed an individual feels to his or her
Peplau, & Hill, 1996; Drigotas, Rusbult, & Verette, 1999; Rusbult, partner in a romantic involvement, the higher is the likelihood that
1983; Rusbult & Buunk, 1993; Rusbult & Martz, 1995; Wie- this relationship will persist over time (e.g., Drigotas & Rusbult,
selquist, Rusbult, Foster, & Agnew, 1999; for an overview, see 1992; Lund, 1985; Rusbult, 1983; Rusbult & Martz, 1995). Fur-
thermore, commitment also produces behavior that extends the
longevity of a relationship (e.g., accommodation; Rusbult, Verette,
Elisabeth Frank and Veronika Brandstätter, Institute of Psychology,
Whitney, Slovik, & Lipkus, 1991; Wieselquist et al., 1999; dero-
University of Munich, Munich, Germany. gation of alternative partners; D. Johnson & Rusbult, 1989; Lydon,
Part of this article was presented at the 12th general meeting of the Meana, Sepinwall, Richards, & Mayman, 1999; Miller, 1997;
European Association of Experimental Social Psychology, Oxford, En- Simpson, Gangestad, & Lerma, 1990; willingness to sacrifice; Van
gland, July 1999. This research was supported by German Science Foun- Lange et al., 1997). In summary, there is a lot of evidence to
dation Grant Fr 472/16-1 to Veronika Brandstätter. We are thankful to suggest that the degree of commitment affects the duration of the
Hermann Brandstätter at the University of Linz, Austria, who gave us the relationship.
opportunity to collect data for the present research question within his Aside from assessing the extent of the commitment, however,
larger research project on personality and close relationships. We also
one might also analyze the specific content of a given commitment
thank Claudia Ehrenhuber and Marketa Hanetslegrova for their assistance
in data collection. We gratefully acknowledge the helpful comments of
and ask why a person feels committed to his or her romantic
Dieter Frey, Günter W. Maier, and Stefan Schulz-Hardt. partner. That is, one could try to specify the incentives a person
Correspondence concerning this article should be addressed to Elisa- strives for in his or her relationship. From a motivational point of
beth Frank, Institute of Psychology, University of Munich, Leopold- view, analyzing commitment in terms of its incentives permits a
strasse 13, 80802 Munich, Germany. E-mail: [email protected] deeper understanding of underlying affective processes (e.g.,
Journal of Personality and Social Psychology, 2002, Vol. 82, No. 2, 208 –221
Copyright 2002 by the American Psychological Association, Inc. 0022-3514/02/$5.00 DOI: 10.1037//0022-3514.82.2.208

208
 APPROACH VERSUS AVOIDANCE COMMITMENT 209

Heckhausen, 1991; Higgins, 1998). This, in turn, would be useful Such content aspects of commitment have received some atten-
in going beyond stability concerns and investigating affective tion in the organizational domain. For example, Meyer et al.
effects of commitment that are related to the quality of a relation- (1993) reported differential correlations between these different
ship—another core dimension of relationship functioning along- types of commitment and indices of goal-directed behavior with
side stability. respect to occupational commitment to nursing (e.g., the intention
Although in the investment model various factors contributing to stay in the nursing profession, absenteeism). Furthermore, their
to an individual’s commitment to his or her partner are differen- research has shown attitudes and affective variables such as work
tiated (i.e., satisfaction, alternatives, investments), the effects of satisfaction to be differentially related to these distinct commit-
these variables are thought to add up to determine the degree of ment types (see also Becker, 1992). Obviously, the explanatory
commitment to the partner. As a consequence, this summary power of different bases for commitment extends from behavior-
conception of commitment does not allow for an analysis of the related variables into the field of emotional aspects.
specific content of a given commitment. However, some recent However, whereas the organizational domain has benefited from
approaches to commitment have started to support a finer grained differentiating the concept of commitment, the field of intimate
analysis of the content of a given commitment (e.g., Brickman, relationships is still waiting for empirical work to be done in this
1987; M. Johnson, 1991; Lydon, 1996; Lydon, Pierce, & O’Regan, area. One exception is Lydon et al.’s (1997) study of the effect of
1997; for related approaches outside the relationship domain, see different types of commitment on relationship satisfaction and
Becker, 1992; Matthieu & Zajac, 1990; Meyer & Allen, 1991; coping with relationship dissolution in terms of affect and illness
Meyer, Allen, & Smith, 1993; Novacek & Lazarus, 1990; Somers, symptoms. In their study, moral commitment (as defined by M.
1995). Johnson, 1991) predicted distress created by the ending of a
romantic relationship, whereas enthusiastic commitment, which
The Content Approach to Commitment parallels Johnson’s personal commitment, had no effect on affec-
tive variables whatsoever.
For example, both M. Johnson (1991), in the relationship do- Aside from the meager empirical basis, what is even more
main, and Meyer and Allen (1991), with respect to organizational important is that in M. Johnson’s (1991) but also in Meyer et al.’s
commitment, have argued that there are three different themes that (1993) conceptualization, it remains unclear on a theoretical level
need to be distinguished in the definition of commitment: (a) a exactly what the underlying, discriminating dimensions of the
“want to,” (b) an “ought to,” and (c) a “have to” type of commit- various proposed types of commitment are.
ment. M. Johnson (1991) named these types of commitment (a)
personal, (b) moral, and (c) structural commitment, whereas A Motivational Approach to Commitment Dimensions:
Meyer and Allen (1991) called them (a) affective, (b) normative, Approach and Avoidance
and (c) continuance commitment.
We take the view that only by filling this theoretical gap with a
Personal–Affective Commitment genuinely motivational concept does a sound analysis of different
commitment dimensions and an explanation for their distinct ef-
By personal commitment, M. Johnson (1991) was referring to fects seem possible. We propose to apply the fundamental distinc-
“the sense of wanting to continue a relationship” (p. 12) that results tion drawn in motivation psychology between approach and avoid-
from a positive attitude toward the partner and the relationship and ance motivation; that is, behavior directed at approaching positive
from relational identity (cf. Meyer & Allen’s, 1991, positive af- incentives and avoiding negative incentives, respectively (e.g.,
fective attachment). Atkinson, 1957; Elliot & Church, 1997; Heckhausen, 1991; Hig-
gins, 1998). From this point of view, one might feel committed to
Moral–Normative Commitment one’s romantic partner because one strives for the positive incen-
tives associated with continuing the relationship (i.e., approach
M. Johnson (1991) conceived of moral commitment as “the commitment). In the same vein, one might feel committed to one’s
feeling that one ought to continue the relationship. . . . [the feeling romantic partner because one tries to avoid the negative incentives
that] ‘I am not doing what I want to, but rather what I feel is right’” associated with breaking up the relationship (i.e., avoidance
(p. 121), which involves a strong sense of self-constraint and commitment).
results from person-specific obligation. This notion of obligation is The rationale for differentiating types of commitment as pro-
also part of Meyer and Allen’s (1991) normative commitment posed by M. Johnson (1991) or Meyer and Allen (1991) would
component. hence be based on the distinction between approach and avoidance
motivation: Whereas the personal–affective (i.e., “want to”) com-
Structural–Continuance Commitment mitment component reflects an approach orientation by referring
to positive incentives (e.g., positive affection, relational identity)
Finally, M. Johnson’s (1991) structural and Meyer and Allen’s that can be achieved by maintaining a relationship, the moral–
(1991) continuance commitment reflect “the feeling that one has to normative (i.e., “ought to”) and structural– continuance (i.e., “have
continue the relationship” (M. Johnson, 1991, p. 122) and is due to to”) components represent an avoidance orientation by involving
“irretrievable investments, [negative] social reaction [to relation- negative incentives. According to Higgins’s (1998) regulatory
ship dissolution], difficulty of termination procedures, and [lack focus theory, duties, obligations, and responsibilities that are de-
of] availability of acceptable alternatives” (M. Johnson, 1991, p. fined by the pressure to live up to one’s own sense of right and
122). wrong or to obligations toward particular persons represent ought
210 FRANK AND BRANDSTÄTTER

self-guides.1 Ought self-guides “are goals that a person must attain Values represent “trans-situational goals that serve as guiding
or standards that must be met. . . . Discrepancies to such goals principles in the life of a person” (Schwartz, 1995, p. 665). They
represent the presence of negative outcomes” (Higgins, 1998, p. are represented on an abstract level in the hierarchy of self-
5). Ought regulation is inherently associated with an inclination regulatory variables and influence a person’s cognitive–affective
toward avoiding breaches of duty as a self-regulatory strategy appraisal of more concrete, situation-specific behavioral options
(e.g., Higgins, Roney, Crowe, & Hymes, 1994). Moreover, the (Carver & Scheier, 1990; Feather, 1990)—in one’s ongoing rela-
structural– continuance (i.e., “have to”) component, as specified by tionship, for example. The great variety of personal values can be
M. Johnson or Meyer and Allen, refers to negative outcomes that classified into personal value domains (Schwartz & Sagiv, 1995).
are connected to a dissolution of the relationship (e.g., negative Some of these domains represent what Higgins (1998) has termed
social reactions, difficulty of termination procedure) and thus also a promotion focus in his regulatory focus theory. According to
represents avoidance commitment. Higgins (1998), “a promotion focus is concerned with accomplish-
ments, hopes, and aspirations” (p. 16) and is linked to a sensitivity
Differential Effects of Approach Versus Avoidance to positive incentives and approach as a strategic means. Some
Motivation on Cognition, Affect, and Behavior other value domains express values that characterize a prevention
focus in Higgins’s (1998) terminology; this prevention focus is
In the approach–avoidance literature (e.g., regulatory focus the- “concerned with safety, responsibilities, and obligations” (p. 16)
ory; Higgins, 1998), a well-documented phenomenon is that ap- and is associated with the sensitivity to negative incentives and
proaching positive incentives as opposed to avoiding negative avoidance as a strategic means.
incentives has distinct effects on cognitive, affective, and behav- Following this line of reasoning, we hypothesized that promo-
ioral processes in goal striving (e.g., Elliot & Church, 1997; Elliot tion focus values (e.g., universalism, humanism) would show a
& Sheldon, 1997; Higgins, 1998; Higgins & Tykocinski, 1992; positive correlation with approach commitment but not with avoid-
Schmalt, 1999). For example, when an individual is guided by the ance commitment. In the same vein, we expected prevention focus
fear of aversive possibilities, the prevailing focus on negative values (e.g., security, conformity, tradition) to show a positive
outcomes and negative information is likely to evoke threat ap- correlation with avoidance commitment but not with approach
praisals, anxiety, and self-protection processes (Elliot & Sheldon,
commitment (Hypothesis 1a).
1997; Higgins, 1998). This, in turn, lessens the extent to which
A second differential correlate of approach versus avoidance
goal pursuit is experienced as being enjoyable and fulfilling (Elliot
commitment refers to the similarity between partners with respect
& Harackiewicz, 1996; Elliot & Sheldon, 1997). Furthermore,
to relevant beliefs. As Hendrick and Hendrick (1992) put it, “the
avoidance goals are associated with more physical symptoms than
crucial factor in attraction may be . . . the relative similarity or
are approach goals (Elliot & Sheldon, 1998; Emmons & Kaiser,
difference between self and the other” (p. 25) on beliefs and
1996) and— by decreasing perceived competence—result in lower
attitudes. Although the similarity-attraction hypothesis has not
self-esteem and fewer feelings of personal control (Elliot & Shel-
remained unchallenged (Rosenbaum, 1986; see also Byrne, Clore,
don, 1997). Most important in the context of our study, avoidance
& Smeaton, 1986), research suggests that partners in romantic
motivation affects psychological adjustment outcome variables;
relationships “who agree in their definition of their relationship are
for instance, by impairing subjective well-being (Elliot & Sheldon,
1997; Elliot, Sheldon, & Church, 1997), increasing negative emo- more apt to escalate their commitment” (Huston & Levinger, 1978,
tionality (Roney, Higgins, & Shah, 1995), and lowering life satis- p. 142). Because mutual understanding and interpersonal attraction
faction (Elliot & Sheldon, 1997). constitute positive incentives in a relationship, we hypothesized
that approach commitment but not avoidance commitment is pos-
itively correlated with the similarity between partners’ views about
Goals of the Present Study a good relationship (Hypothesis 1b).
M. Johnson’s (1991) as well as Meyer and Allen’s (1991) A third and final correlate of avoidance commitment, as distinct
presentations of commitment as a tripartite construct are intellec- from approach commitment, relates to previous investments in the
tually inspiring accounts of commitment processes. However, their relationship in terms of resources (e.g., time and money) that
differentiation of commitment types seems somewhat arbitrary, would be irretrievably lost if the partners ended the relationship.
lacking a sound theoretical rationale. This theoretical basis could “Investments of [this] sort intensify commitment by increasing the
be derived from motivation theory, which permits predictions costs of ending the relationship” (Rusbult, 1991, p. 158). In ac-
about differential effects of approach versus avoidance regulation cordance with a dissonance explanation of entrapment– escalation
on psychological functioning. of commitment (Brockner, 1992) and the phenomenon of sunk
To evaluate the validity of our approach–avoidance distinction, costs (Arkes & Blumer, 1985), a person who has invested heavily
we sought to identify different correlates of each type of commit-
ment providing convergent and discriminant validity for the pos- 1
tulated distinction. Our predictions were based on different strands The decisive factor for classifying an instance as ideal versus ought
regulated is whether the psychological situation involves the presence of
of research: (a) personal values as guidelines for life (Carver &
positive or negative outcomes, respectively. For example, values that one
Scheier, 1990; Schwartz & Sagiv, 1995), (b) the similarity between wholeheartedly embraces (e.g., being a faithful partner because one vowed
partners regarding their views about a good relationship (e.g., faithfulness) do not represent ought self-guides but rather ideal self-guides
Antill, 1983; Byrne, 1971; Hassebrauck, 1996; Levinger, 1979), (Higgins, 1998), that is, an approach and not an avoidance motivation,
and (c) relationship duration as an indicator of irretrievable invest- because living up to one’s values involves the presence of positive out-
ments (M. Johnson, 1991; Rusbult, 1980). comes (e.g., contentment).
 APPROACH VERSUS AVOIDANCE COMMITMENT 211

in a relationship should be less likely to leave the relationship months. The first testing period (Time 1) took place in April. Subsequent
because he or she wants to avoid the negative experience of losing testing periods occurred 6 months (Time 2, October) and 13 months
the investments on the dissolution of the relationship. Following (Time 3, May of the following year) after the first testing period.
this line of thought, one should find that investments in terms of In each testing period, participants were contacted by student experi-
menters. Questionnaires including our main variables were answered in the
the duration of a relationship correspond positively with avoidance
presence of a student experimenter and independently of the romantic
commitment but not with approach commitment (Hypothesis 1c).
partner. Each participant was instructed to assign himself or herself a
Besides demonstrating the validity of our differentiation of two personal code that was to be used across all three testing periods to identify
commitment types, our even more important research objective participants without infringing on their anonymity. The mean age of the
was to examine the predictive value of approach versus avoidance sample was 25 years, with a range of 16 to 36 years. For a variety of
commitment in explaining relationship quality variables in a lon- reasons, the sample size dropped from 67 couples to 54 couples at Time 2
gitudinal design. (81% of the original sample) and 43 couples at Time 3 (64% of the original
One important aspect of relationship quality is the partners’ sample). Four couples moved to another city, 2 withdrew from the study
overall satisfaction with the relationship (Berscheid, 1994; Hasse- because their relationship had broken up, 2 failed to remember their
brauck, 1991; Hendrick, 1988; for an overview, see Glenn, 1990). identifying code, and 3 explicitly declined to continue their participation in
the study. For the remaining 13 couples, we have no information about why
Other measures of relationship quality are the percentage of time
they did not answer the second and/or third questionnaire. Those couples
that one experiences well-being in the presence of one’s partner,
who withdrew after the first or second testing period did not differ signif-
on the one hand, and the frequency of positive emotions, on the icantly from those who remained in the sample throughout Times 1–3 in
other hand. terms of any of our measures at Time 1.
On the basis of the evidence from the approach–avoidance
literature that avoidance-based motivation increases negative emo- Questionnaires
tionality as well as impairs subjective well-being and life satisfac-
tion, whereas approach-based motivation increases positive emo- Personal values. Schwartz and Bilsky’s (1990) value list was admin-
tionality as well as furthers well-being and life satisfaction (e.g., istered to participants at Time 1. This list contains 58 values derived
from 10 value domains (see Schwartz & Sagiv, 1995). Participants had to
Elliot & Sheldon, 1997; Elliot et al., 1997; Roney et al., 1995), we
rate each value according to its importance as a guideline for their own life.
formulated a second set of hypotheses. We expected that in the Judgments were made on a 9-point scale ranging from ⫺1 (contrary to my
long run, the approach form of commitment would be positively values) to 7 (extremely important). We created an index of promotion-
associated with relationship satisfaction, whereas the avoidance related values (Higgins, 1998) by averaging the scores on the universalism
form of commitment would be negatively associated with relation- (including, e.g., “inner harmony,” “tolerant”) and humanism (e.g., “a sense
ship satisfaction (Hypothesis 2a). We predicted the same pattern in life,” “true friendship”) value domains (Cronbach’s ␣ ⫽ .78).2 Analo-
for emotionality in terms of the percentage of time that one gously, we constructed a composite score comprising value domains re-
experiences well-being in the presence of one’s partner (Hypoth- ferring to the prevention of negative outcomes (Higgins, 1998), that is,
esis 2b) and the frequency of positive emotions (Hypothesis 2c). security (e.g., “familiar security,” “clean”), conformity (e.g., “politeness,”
To sum up, we first aimed at demonstrating that M. Johnson’s “self-discipline”), and tradition (e.g., “respect towards tradition,” “moder-
ate”). Cronbach’s alpha was .80. The remaining dimensions (e.g., “power,”
(1991) and Meyer and Allen’s (1991) different types of commit-
“performance”) were not analyzed any further.
ment can theoretically be integrated into an approach–avoidance Similarity between partners regarding ideas about a good relationship.
perspective. Next, we sought to validate these two dimensions by The definition of what constitutes a good relationship affects one’s attitude
a different pattern of correlates. Finally, we examined the contri- toward one’s actual relationship. We assessed participants’ conceptions of
bution of approach and avoidance commitment in predicting rela- what characterizes a good relationship at Time 1 using a 32-item scale
tionship quality variables over the course of time. constructed by Hassebrauck (1996). Participants rated the extent to which
each item was important for a good relationship on a scale ranging from 0
(not at all important) to 6 (extremely important). Sample items are “accept
Method
each other,” “share similar views,” and “have fun together.” Items were
Overview later averaged to produce a single measure for ideas about a good rela-
tionship (Cronbach’s ␣ ⫽ .90). We assessed similarity between partners
To test our hypotheses, we used a longitudinal design. Members of with respect to ideas about what constitutes a good relationship by first
couples separately answered three questionnaires about their relationship calculating the absolute difference between the 32 z-transformed ideas of
over a period of 13 months. All Time 1–Time 3 questionnaires included the 2 members of the couple and then averaging these differences across
measures referring to (a) descriptive data such as relationship status (e.g., the 32 items of the scale. As a result, lower numbers indicate higher
married or not, living together or not) and duration of the relationship so similarity between partners.
far, (b) commitment, (c) relationship satisfaction, and (d) emotions expe- Commitment. To our knowledge, M. Johnson (1991) has not developed
rienced while with the partner. In addition, personal values and similarity an instrument to test his commitment types, whereas Meyer and colleagues
regarding ideas about what characterizes a good relationship were assessed have (see Meyer et al., 1993), although not for the relationship but for the
at Time 1. organizational domain. We therefore adapted items used by Meyer et al.
(1993) for our purposes. For the sake of coherence, we stick to Meyer et
Participants and Design al’s. (1993) terminology of affective, normative, and continuance commit-
ment in this article. Drawing on this measure, we used affective commit-
Sixty-seven heterosexual couples, recruited among their acquaintances
by 15 social science students of a research seminar on close relationships
2
at the University of Linz, Austria, participated in the study. Only couples Because all analyses are reported on the dyad level in the Results
who had been seriously involved for at least 6 months were invited to section, reliability analyses have been performed with the dyad average as
participate. Data were collected at three testing periods over a period of 13 well.
212 FRANK AND BRANDSTÄTTER

ment variables to represent approach commitment, whereas avoidance On average, at Time 1 couples had been involved for 42.6
commitment was supposed to be captured by normative and continuance months, with a range between 6 months and 10 years (SD ⫽ 29.7
commitment items. months). Thirty-one couples were living together, and 36 couples
Commitment measures consisted of a series of 12 commitment items3 were not. Among those not living together, the frequency of dating
and were administered to participants at all three testing periods. Nine of
was rather high (M ⫽ 5.1 days per week at Time 1, M ⫽ 5.4 at
these items were based on an adaptation of the Meyer et al. (1993) items
Time 2, and M ⫽ 4.9 at Time 3). At Time 1, only 1 couple reported
for the relationship domain (in the questionnaire, items were presented at
random). The items for affective commitment were (a) “I regret having that they were married, but during the course of the study another 5
entered this relationship” (reverse coded), (b) “I am attached to my part- couples got married. Eighteen couples had been separated once
ner,” (c) “I identify with my partner,” and (d) “I would not suffer a lot if before; 49 had not. Of the 41 couples answering the question of
this relationship would break up” (reverse coded). The items for normative whether they had children together or not, 8 answered yes, and 33
commitment were (a) “Even if it were to my advantage, I do not feel it answered no.
would be right to end this relationship,” (b) “I feel a responsibility towards
my partner to continue this relationship,” and (c) “People who are impor-
tant to me would react negatively if I were to end this relationship.” The Strategy for Analysis
items for continuance commitment were (a) “Too much of my life would
be disrupted if I were to end this relationship,” and (b) “I have put so much Data for partners in a given relationship are not statistically
into this relationship that ending it would be very painful.” independent (cf., e.g., Gonzalez & Griffin, 1997; Judd, McClel-
Another 3 items—those that could be applied in the relationship con- land, & Culhane, 1995; Kenny, 1996). Approaches accounting for
text—were taken from Brunstein’s (1993) 6-item commitment scale and this problem, such as hierarchical linear modeling (Bryk & Rau-
were not assigned a priori to a particular commitment type. These items are denbush, 1992) or Gonzalez and Griffin’s (1997) pairwise corre-
(a) “Even if it means a lot of effort I’ll do everything necessary to preserve lation method, are not applicable to our data, either because a
this relationship,” (b) “No matter what happens, I will not give up this larger sample than ours is required or because there are no rec-
relationship,” and (c) “I sometimes doubt if I should continue this rela- ommendations for the types of analyses performed by us (e.g.,
tionship” (reverse coded). All 12 items had to be rated by participants
factor analysis). Consequently, we decided to follow the more
according to the extent to which each statement was true for themselves
conservative—albeit power-reducing—strategy of using average
(1 ⫽ not at all true for me, 5 ⫽ very much true for me).
Relationship satisfaction. To assess relationship satisfaction, we dis- scores on a couple level. These results are largely consistent with
tributed the German version of Susan Hendrick’s (1988) Relationship analyses performed on the individual level and those run sepa-
Assessment Scale (Hassebrauck, 1991) at each testing period (Times 1 rately for men and women.
through 3). Participants were told to rate the seven items of the scale
according to their present situation. Ratings were made on 7-point scales
ranging from 1 (very) to 7 (not at all). Sample items are “How well does
Commitment Dimensions
your partner fulfill your needs?” and “How good is your relationship
Commitment items from Times 1–3 were subjected to three
compared with most other couples’ relationships?” After reversing the
separate principal-components analyses. For each analysis, four
scale, we averaged the ratings across all items to generate an index of
relationship satisfaction (Cronbach’s alpha for the three testing periods ⫽ factors with eigenvalues greater than 1 emerged. The initial ei-
.90, .90, and .93, respectively). genvalues of the unrotated solutions were 2.9, 2.3, 1.2, and 1.0
Emotions while with partner. The emotions participants experienced (Time 1); 2.9, 1.9, 1.2, and 1.1 (Time 2); and 2.9, 2.5, 1.4, and 1.3
while in the company of their romantic partner were assessed using two (Time 3). Because all scree tests revealed a clear elbow after two
different measures. First, general well-being was assessed in quantitative factors, we extracted two factors at each testing period. Eigenval-
terms. Participants had to indicate the percentage of time during the past 2 ues and factor loadings after varimax rotation are presented in
weeks they had felt subjectively well when their partner was present. Table 1.
Secondly, the quality of the emotions they experienced while in the Notably, results from the factor analyses are, in general, in line
company of their romantic partner was assessed using a list of three
with our theoretical a priori assignment of items to commitment
positive emotional nouns (i.e., happiness, love, sexual satisfaction) and
dimensions. Items from the first factor (see Table 1) refer for the
four negative emotional nouns (i.e., sadness, anger, disappointment, con-
tempt). Participants indicated the extent to which they felt these moods most part to affective attachment and, thus, affectional rewards
during the past 2 weeks while in the company of their romantic partner associated with the relationship itself, which represent approach-
(1 ⫽ never, 5 ⫽ always). After recoding negatively worded nouns, we related commitment. Therefore, we interpret this factor as ap-
constructed an overall composite score of positive affect (Cronbach’s alpha proach commitment. Most of the items in Factor 2 refer to costs
for the three testing periods ⫽ .80, .80, and .88, respectively). associated with the termination of the relationship and the obliga-
tion to remain in the relationship—all negative incentives associ-
Results ated with leaving the relationship. Therefore, we interpret this
Descriptive Data second factor as avoidance-related commitment. Both factors ac-
count for approximately equal shares of variance in the data over
Twenty-four (17.9%) participants had a Hauptschule (a second- the three testing periods (see Table 1).
ary school leaving certificate), 21 (15.7%) had a Fachschule (a A closer look at the factor loadings reveals that Factor 2 in-
technical college leaving certificate), 70 (52.2%) had a Matura (a volves doubt about continuing the relationship, stemming from
school-leaving examination at grammar school needed for entry to Brunstein (1993), and Factor 1 comprises willingness and deter-
higher education), and 18 (13.4%) had a university diploma.
Within couples, 37 couples had the same education level, 29
couples differed with respect to education level reached, and 1 3
Because commitment items were later subject to separate factor anal-
couple was not classifiable. yses for each testing period, we do not report Cronbach’s alphas here.
 APPROACH VERSUS AVOIDANCE COMMITMENT 213

Table 1
Varimax-Rotated Principal Components of Commitment Times 1–3

Time 1 Time 2 Time 3

Item F1 F2 F1 F2 F1 F2

Approach commitment
I sometimes doubt if I should continue this relationship. (reverse coded) .85 .59 .78
I am attached to my partner. .66 .67 .37 .67
I regret having entered this relationship. (reverse coded) .64 .62 .79
I would not suffer a lot if this relationship would break up. (reverse coded) .64 .35 .40
I identify with my partner. .59 .57 .59
Avoidance commitment
Even if it were to my advantage, I do not feel it would be right to end this relationship. .75 .69 ⫺.41 .62
Too much of my life would be disrupted if I were to end this relationship. .65 .54 .64
I feel a responsibility towards my partner to continue this relationship. .64 ⫺.37 .53 .70
Even if it means a lot of effort I’ll do everything necessary to preserve this relationship. .56 .66 .82
No matter what happens, I will not give up this relationship. .47 .52 .64 .60
People who are important to me would react negatively, if I were to end this
relationship. .50 .56 .39 .44
I have put so much into this relationship that ending it would be very painful. .49 .67 .27
Eigenvalue of factor 2.7 2.6 2.9 2.0 2.7 2.7
% variance explained by factor 22.8 21.5 24.1 16.3 22.6 22.4

Note. With one exception, only factor loadings greater than .35 are reported. F1 is the approach commitment factor; F2 is the avoidance commitment
factor. F ⫽ factor.

mination. The avoidance item “I have put so much into this relation- tively with approach commitment, albeit not reaching conventional
ship that ending it would be very painful” was only loosely attached significance, p ⬍ .07 (one-tailed), but not with avoidance commit-
to the avoidance factor (Factor 2) at Time 3. However, because the ment (see Table 2). More important, the two correlations differed
avoidance factor loadings for Time 1 and Time 2 were satisfactorily significantly from each other. Using a formula put forward by Cohen
high and also because it corresponds to our a priori assignment to and Cohen (1983, p. 57) resulted in t(64) ⫽ 1.69 for the difference
commitment types, we left this item in the avoidance factor. A similar between correlations, p ⬍ .05 (one-tailed). In line with predictions,
logic applies to the items “No matter what happens, I will not give up prevention focus values (i.e., security, conformity, and tradition) cor-
this relationship” and “People who are important to me would react related significantly positively with avoidance commitment but not
negatively if I were to end this relationship,” both of which we left with approach commitment. Again, the difference in correlation co-
with the avoidance factor (Factor 2). efficients proved significant, t(64) ⫽ 1.68, p ⬍ .05 (one-tailed).
For the subsequent analyses, after reversing negatively worded Similarity on ideas about a good relationship (Hypothesis 1b).
items, we calculated composite scores for the two commitment As can be seen in Table 2, our hypothesis regarding the association
dimensions—approach and avoidance—at the three measurement between commitment types and similarity between partners was con-
periods on the basis of the unweighted mean of the responses to the firmed, with a high similarity on central ideas about a good relation-
items pertaining to each factor (means and standard deviations of ship corresponding significantly to higher approach commitment—
these scores are shown in Table 2). As assessed by coefficient but not avoidance commitment. Again, this difference in correlation
alpha, the reliabilities of the two commitment dimensions appear coefficients proved significant, t(64) ⫽ 2.67, p ⬍ .01 (one-tailed).4
not to have been too bad (with one exception, all ␣s ⬎ .69; cf. Relationship duration (Hypothesis 1c). The longer an involve-
Table 2). Stability between testing periods (see Table 2) was ment persists, the more has been invested—meaning that the cost of
acceptably high for both commitment dimensions (all rs ⬎ .53), leaving the relationship increases with the duration of the relationship.
considering the length of the interval between testing periods.
4
We are well aware that the use of difference scores is not unproblem-
Summary Statistics atic (e.g., Griffin, Murray, & Gonzalez, 1999). To circumvent one of the
The means and standard deviations of all other main study problems associated with it—the problem of different variances in the two
variables as well as their correlations with the commitment mea- variables that compose the difference score—we used z-transformed mea-
sures are given in the lower part of Table 2. The intercorrelations sures. However, we also pursued other strategies to test our assumption.
For example, we computed correlation coefficients for each couple on the
between major study variables— except for approach and avoid-
basis of a transposed data matrix, treating the 2 members of each couple as
ance commitment (see Table 2)—are given in Table 3. items and the 32 items of the scale as cases. Fisher’s z transformation was
then used to make correlation coefficients comparable. As expected, the
Correlates of Approach Versus Avoidance Commitment correlation between these coefficients and approach commitment at Time 1
(Hypotheses 1a–1c) yielded r ⫽ .23, p ⬍ .04 (one-tailed). Following another procedure, applied
by Murray, Holmes, and Griffin (1996), we also categorized the 2 mem-
Personal values (Hypothesis 1a). As expected, the promotion bers’ scores on the 32 ideas of a good relationship separately in low or high
focus scale comprising universalism and humanism correlated posi- importance and then examined the resulting 2 ⫻ 2 table in an ANOVA
214 FRANK AND BRANDSTÄTTER

Table 2
Means, Standards Deviations, and Correlation Coefficients for Approach (AP) and Avoidance (AV) Commitment Scales
and Other Major Study Variables

Variable M SD 1 2 3 4 5 6

1. AP, T1 4.22 0.46 (.74)

2. AP, T2 4.30 0.38 .65*** (.56)
3. AP, T3 4.18 0.54 .53*** .62*** (.69)
4. AV, T1 2.92 0.57 .15 ⫺.04 ⫺.11 (.69)
5. AV, T2 3.04 0.62 .36** .03 .00 .65*** (.73)
6. AV, T3 3.01 0.55 .15 .00 .07 .54*** .81*** (.71)
7. Promotion focus valuesa 4.64 0.60 .19 .05 ⫺.03 ⫺.08 ⫺.09 ⫺.28
8. Prevention focus values 3.33 0.69 .15 .20 ⫺.01 .40*** .40** .22
9. Similarityb 0.94 0.42 ⫺.33** ⫺.35** ⫺.03 .08 .10 .05
10. Investments (duration of relationship in months)c 42.90 30.30 ⫺.01 ⫺.12 ⫺.04 .26* .20 .20
11. RELSAT, T1 4.73 0.67 .84*** .48*** .49*** .07 .29* .17
12. RELSAT, T2 4.80 0.73 .67*** .69*** .52*** ⫺.12 .10 .06
13. RELSAT, T3 4.51 0.92 .59*** .62*** .88*** ⫺.16 ⫺.06 .03
14. WELL, T1 79.89 14.10 .64*** .38** .51** ⫺.08 ⫺.09 ⫺.16
15. WELL, T2 80.32 19.10 .44** .53*** .38* ⫺.36* ⫺.09 ⫺.08
16. WELL, T3 77.51 20.20 .32* .41** .78*** ⫺.33* ⫺.16 ⫺.05
17. POSEM, T1 3.68 0.30 .55*** .33* .43** .01 .11 .05
18. POSEM, T2 3.63 0.36 .44** .55*** .30* ⫺.19 .00 ⫺.09
19. POSEM, T3 3.52 0.40 .28 .41** .68*** ⫺.05 .05 .02

Note. Data in parentheses are reliabilities estimated using the coefficient alpha. T ⫽ time; RELSAT ⫽ relationship satisfaction, measured on a scale ranging
from 1 to 7, with higher values indicating more satisfaction; WELL ⫽ percentage of time that a participant experiences well-being when with partner; POSEM ⫽
frequency of positive (and reversed negative) emotions, measured on a scale ranging from 1 to 5, with higher values indicating more positive emotions.
a
Personal values and similarity between partners regarding ideas about a good relationship were assessed only at Time 1. b This measure is based on the
difference between men’s and women’s standardized scores. Lower values represent higher similarity. c Note that duration cannot vary from Time 1 to
Time 3—aside from variation due to missing data— because testing periods Time 2 and Time 3 occurred after a fixed interval. Because logarithmic values
are difficult to interpret, the mean and standard deviation are reported for the raw values; however, all correlations are based on the log of the duration.
* p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

This hypothesis is confirmed by the positive correlation between the pany of the romantic partner, whereas avoidance commitment is
log of the duration5 and avoidance commitment (see Table 2), negatively associated with them, we performed a series of hierar-
whereas there was no significant correlation with approach commit- chical regression analyses. In each analysis, we controlled for our
ment. Testing this difference yields t(64) ⫽ 1.72, p ⬍ .05 (one-tailed). baseline measures of relationship satisfaction at Time 1 and emo-
Thus, in sum, the correlational patterns for approach versus tions at Time 1. In a second step, Time n relationship satisfaction
avoidance values, the similarity between partners regarding their and emotionality were each predicted by means of the approach
ideas about a good relationship, and the duration of the relationship and avoidance commitment also assessed at Time 1.
provide convergent and discriminant validity for our distinction Predicting relationship satisfaction (Hypothesis 2a). As
between approach and avoidance commitment. shown at the top of Table 4, when we controlled for relationship
satisfaction at Time 1, approach and avoidance commitment at Time 1
Longitudinal Results: Testing Long-Term Effects of were significant predictors for relationship satisfaction at Time 2 (see
Approach and Avoidance Commitment on Satisfaction and also upper part of Figure 1) and Time 3, respectively.6 More impor-
Emotions (Hypotheses 2a–2c) tant however, and supporting our Hypothesis 2a, approach commit-
To test our second set of hypotheses (Hypotheses 2a–2c), which ment always yielded a positive beta weight for predicting relationship
state that approach commitment is positively associated with re- satisfaction; that is, it was positively associated with relationship
lationship satisfaction and with one’s emotions while in the com- satisfaction 6 and 13 months later, whereas avoidance commitment
received a negative beta weight in both cases.7

approach. If our assumption was correct, then approach commitment 5

should be higher in the high– high and low–low cells than in the other two Because duration was not normally distributed but skewed, we used a
cells. Accordingly, the ANOVA revealed an interaction between the 2 logarithmic transformation for all analyses concerning duration. However, the
partners’ scores, with F(1, 63) ⫽ 2.83, p ⫽ .05 (one-tailed). When we same relationships held true when we used the raw values of duration instead.
6
inspected the table, it was clear that approach commitment was especially Predicting Time 3 relationship satisfaction by Time 2 approach and
high if both partners had given high importance estimates for the 32 ideas avoidance commitment yields similar results.
7
about a good relationship (i.e., were categorized as high). To sum up, all Exploratory analyses of whether avoidance commitment moderates the
these analyses confirm the results of our difference score correlation. relationship between approach commitment and relationship satisfaction—
Because a difference score measure is intuitively plausible as an indicator tested by introducing the interaction between avoidance commitment and
of similarity and because this approach best joins the rest of the analyses approach commitment into the regression equation—yielded no significant
concerning Hypothesis 1, we focus on the difference score results. results.
 APPROACH VERSUS AVOIDANCE COMMITMENT 215

Table 3
Intercorrelation Between Major Study Variables

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

1. Promotion focus values

2. Prevention focus values .04
3. Similarity ⫺.29* ⫺.24*
4. Investments (duration of relationship) ⫺.01 .01 ⫺.05
5. RELSAT, T1 .25* .07 ⫺.27* ⫺.10
6. RELSAT, T2 .20 .12 ⫺.32* ⫺.30* .70***
7. RELSAT, T3 .11 ⫺.09 ⫺.12 ⫺.13 .57*** .73***
8. WELL, T1 .18 ⫺.14 ⫺.22 ⫺.12 .72*** .46** .58***
9. WELL, T2 .16 ⫺.25 ⫺.26 ⫺.20 .46*** .67*** .53*** .44**
10. WELL, T3 .11 ⫺.17 ⫺.07 .02 .29 .50** .85*** .36* .43**
11. POSEM, T1 .24* .00 ⫺.22 ⫺.23 .81*** .61*** .55*** .73*** .39** .25
12. POSEM, T2 .30* .05 ⫺.25 ⫺.36** .52*** .83*** .61*** .42** .73*** .39** .56***
13. POSEM, T3 .05 .04 .02 ⫺.21 .38** .52*** .83*** .47** .34* .76*** .48** .54***

Note. T ⫽ time; RELSAT ⫽ relationship satisfaction, measured on a scale ranging from 1 to 7, with higher values indicating more satisfaction; WELL ⫽
percentage of time that a participant experiences well-being when with partner; POSEM ⫽ frequency of positive (and reversed negative) emotions,
measured on a scale ranging from 1 to 5, with higher values indicating more positive emotions.
* p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

Predicting percentage of time during which well-being is expe- see upper part of Table 5). Obviously, if anything, only the
rienced (Hypothesis 2b). After the term for initial well-being at partner’s approach commitment was significantly associated with
Time 1 had first been partialed out, approach commitment had—as the individual’s satisfaction, indicating that one’s satisfaction
expected—a significantly positive link with the percentage of comes hand in hand with one’s partner’s approach commitment.
positive time spent with the partner, whereas for avoidance com- Second, we investigated lagged associations between one mem-
mitment the relationship was negative (cf. respective beta weights ber’s score on relationship satisfaction at Time n, controlling for
in middle of Table 4). A similar pattern emerged with respect to his or her relationship satisfaction at Time n ⫺ 1 as well as the
predicting the percentage of time when well-being is experienced partner’s score on each commitment type at Time n ⫺ 1 (for
at Time 3 (see also middle of Table 4); however, this time predicting Time 2 and Time 3 satisfaction separately). This strat-
avoidance commitment only displayed a trend in the expected egy only yielded highly significant results for one’s own relation-
direction. Thus, in sum, Hypotheses 2b was fully confirmed with ship satisfaction at Time n ⫺ 1, whereas neither approach nor
respect to Time 2 dependent measures, and partial support was avoidance commitment of one’s partner ever approached statistical
found at Time 3. significance (all ts ⬍ 1). A comparison of the correlation coeffi-
Predicting the frequency of positive emotions (Hypothesis 2c). cients (Cohen & Cohen, 1983) confirmed the conclusion that one’s
When we controlled for the frequency of specific emotions par- own satisfaction at an earlier point in time was a much stronger
ticipants experienced while in the company of their partner at predictor for one’s later satisfaction than was the partner’s ap-
Time 1 in a first step, in line with Hypothesis 2c, we found that proach or avoidance commitment (because all correlation compar-
approach commitment went hand in hand with significantly more isons were significant; all ts ⬎ 4.22, all ps ⬍ .05).
frequent positive emotions, whereas avoidance commitment was We also tested the reverse direction, that is, that one’s approach
associated with less frequent positive emotions at Time 2 (see and avoidance commitment is associated with one’s partner’s
lower part of Table 4). For predicting Time 3 positive emotions relationship satisfaction in the two ways described above. The first
(see also lower part of Table 4), no significant effects of either way, regressing the approach commitment (and avoidance com-
approach commitment or avoidance commitment were found. In mitment, respectively) of one partner on the concurrent relation-
summary, then, Hypothesis 2c was fully supported for predicting ship satisfaction of the other partner again demonstrated associa-
the Time 2 frequency of emotions; however, no support was found tions between approach commitment and relationship satisfaction
for Time 3 frequency of emotions. (see lower part of Table 5). The second strategy (lagged associa-
tions) did not result in any significant proportion of variance
Further Analyses: Mutual Influences Between Partners explained by relationship satisfaction.

To examine interdependence between partners—that is, mutual

Discussion
influences on approach and avoidance commitment and relation-
ship satisfaction—we adopted a mutual cyclical growth model This study examined whether approach versus avoidance com-
approach outlined by Wieselquist et al. (1999). To investigate mitment is the underlying dimension of qualitatively distinct types
whether relationship satisfaction of one dyad member was associ- of commitment identified by M. Johnson (1991) and by Meyer and
ated with approach and avoidance commitment of his or her Allen (1991; Meyer et al., 1993) and attempts to validate this
partner, we first looked at the concurrent associations between an approach–avoidance distinction conceptually. Furthermore, the
individual’s score on relationship satisfaction and his or her part- key aim of the study is to provide evidence for differential asso-
ner’s score on each commitment type (for Times 1–3 separately; ciations between these commitment types, on the one hand, and
216 FRANK AND BRANDSTÄTTER

relationship quality parameters such as satisfaction with the rela-

tionship and positive emotions during times spent with the partner,
on the other hand.

Identifying Approach Versus Avoidance Commitment

Types

The study’s findings with respect to factor analyses run on

commitment items originally stemming from Meyer et al. (1993)
can be summarized as follows: Those commitment themes that
represent positive incentives for the continuation of the relation-
ship—that is, the affective commitment items (which parallel M.

Table 4
Results of Hierarchical Regression Analyses with Approach
(AP) and Avoidance (AV) Commitment

Variables df ␤ adj. R2 F change Figure 1. Path model for predicting relationship satisfaction at Time 2
(upper panel) and path model for predicting approach and avoidance
Hypothesis 2a: Predicting Time 2 RELSAT commitment at Time 2 (lower panel). Data are standardized regression
Step 1: RELSAT, T1 52 .70*** .47 48.48*** coefficients (betas) and correlation coefficients (in parentheses). Unless
Step 2 50 .53 4.05* otherwise stated, all coefficients are significant at p ⬍ .05. For all predic-
AP, T1 .39* tors, regression coefficients are displayed with controls for other predic-
AV, T1 ⫺.20* tors. T ⫽ time; Adj. ⫽ adjusted.

Hypothesis 2a: Predicting Time 3 RELSAT

Step 1: RELSAT, T1 42 .57*** .31 20.06***
Step 2 40 .38 3.44* Johnson’s, 1991, personal commitment type)—represent one fac-
AP, T1 .45* tor, which we have termed approach commitment. Unlike the first
AV, T1 ⫺.24† factor, the second factor, interpreted as avoidance commitment,
mainly comprises normative and continuance commitment items
Hypothesis 2b: Predicting Time 2 WELL
(which cover Johnson’s structural and moral commitment type)—
Step 1: WELL, T1 50 .44 .17 11.71** that is, negative incentives for the termination of the relationship.
Step 2 48 .34 7.38**
Thus, we were able to corroborate our assumption that affective,
AP, T1 .40*
AV, T1 ⫺.40** normative, and continuance commitment from Meyer et al. (1993)
in fact reduce to the two types of commitment we suggested.
Hypothesis 2b: Predicting Time 3 WELL Moreover, the distinct pattern of correlations between commit-
Step 1: WELL, T1 41 .36 .11 5.92* ment types, on the one hand, and approach- versus avoidance-
Step 2 39 .21 3.64* related personal values, similarity between partners’ conceptions
AP, T1 .28 of a good relationship, and investments in the relationship in terms
AV, T1 ⫺.36* of duration, on the other hand, altogether validates the proposed
Hypothesis 2c: Predicting Time 2 POSEM distinction. Approach commitment was more strongly associated
with promotion focus values, as compared with avoidance com-
Step 1: POSEM, T1 52 .56 .30 24.05***
mitment, whereas avoidance commitment but not approach com-
Step 2 50 .40 4.97*
AP, T1 .32* mitment correlated significantly and positively with prevention
AV, T1 ⫺.29* focus values (Hypothesis 1a). The rewarding similarity between
partners on central ideas about how to define a good relationship
Hypothesis 2c: Predicting Time 3 POSEM was positively associated with approach commitment but not with
Step 1: POSEM, T1 42 .40 .21 12.65** avoidance commitment (Hypothesis 1b). By contrast, investments
Step 2 40 .10 ⬍ 1.00 in terms of the duration of the relationship so far only showed a
AP, T1 .12 significant positive correlation with avoidance commitment, not
AV, T1 ⫺.11
with approach commitment (Hypothesis 1c).
Note. adj. ⫽ adjusted; RELSAT ⫽ relationship satisfaction, measured on One objection could be that the two dimensions we obtained
a scale ranging from 1 to 7, with higher values indicating more satisfaction; may be explained alternatively in terms of intrinsic– extrinsic
T ⫽ time; WELL ⫽ percentage of time that a participant experiences motivation (e.g., Ryan & Deci, 2000). Thus, one might argue that
well-being when with partner; POSEM ⫽ frequency of positive (and
reversed negative) emotions, measured on a scale ranging from 1 to 5, with
the approach factor represents an intrinsic form of motivation (e.g.,
higher values indicating more positive emotions. an interest in connecting to the other person), whereas the avoid-
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001. ance factor represents a more extrinsic form of motivation (based
 APPROACH VERSUS AVOIDANCE COMMITMENT 217

Table 5
Predicting Individual Relationship Satisfaction From Partner’s Approach (AP) and
Avoidance (AV) Commitment, and Vice Versa

Women Men

Variable df ␤ R2 F df ␤ R2 F

Predicting IND’s RELSAT on the basis of PART’s AP and AV commitment

IND’s RELSAT, T1 64 .14 6.16** 64 .03 1.96
PART’s AP, T1 .41** .24*
PART’s AV, T1 ⫺.04 .02
IND’s RELSAT, T2 51 .03 ⬍1.00 51 .02 1.42
PART’s AP, T2 .16 .21
PART’s AV, T2 .02 .07
IND’s RELSAT, T3 39 .13 4.17* 39 .07 2.53†
PART’s AP, T3 .41* .32*
PART’s AV, T3 ⫺.29 .11

Predicting IND’s AP commitment on the basis of PART’s RELSAT

IND’s AP, T1 by PART’s RELSAT, T1 65 .24† .04 65 .40** .15
IND’s AP, T2 by PART’s RELSAT, T2 52 .22† .03 52 .16 .03
IND’s AP, T3 by PART’s RELSAT, T3 40 .32* .08 40 .32* .10

Note. IND ⫽ individual; RELSAT ⫽ relationship satisfaction, measured on a scale ranging from 1 to 7, with
higher values indicating more satisfaction; PART ⫽ partner; T ⫽ time.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01.

on gains and losses independent of the affiliative bond). Our explain a significant amount of variance in affective judgment
approach items do indeed have some affinity to an intrinsic part- variables, whereas approach commitment— but not satisfaction—
nership orientation. However, the avoidance items are not explicit should be a significant predictor for motivational variables. Re-
with respect to the intrinsic– extrinsic partition. At least some of gression analyses entering both satisfaction and approach commit-
them allow for an intrinsic interpretation. Empirically, the corre- ment support our hypotheses.8 As expected, only satisfaction, not
lational pattern discussed above speaks against an intrinsic– approach commitment, was a significant predictor for variables
extrinsic interpretation. None of the constructs with which ap- related to affective judgment, such as percentage of well-being
proach or avoidance commitment were correlated lie clearly at one time and frequency of emotions— combined into one single mea-
end of the intrinsic– extrinsic dimension. In addition, on a concep- sure of affect—satisfaction, rpart ⫽ .64, ␤ ⫽ .77, t(67) ⫽ 6.61, p ⬍
tual level, one can quite easily conceive of intrinsic approach
.001; approach commitment, rpart ⫽ .08, ␤ ⫽ .11, t ⬍ 1. It was also
aspects (e.g., affection) as well as of extrinsic approach aspects
the only predictor for a single item tapping resignation; satisfac-
(e.g., family income); in the same vein, one can think of intrinsic
tion, rpart ⫽ ⫺.48, ␤ ⫽ ⫺.68, t(67) ⫽ ⫺4.33, p ⬍ .001; approach
avoidance aspects (e.g., loss of intimacy) as well as of extrinsic
commitment, rpart ⫽ ⫺.07, ␤ ⫽ ⫺.07, t ⬍ 1.
avoidance aspects (e.g., negative social reactions from important
others). In contrast, only approach commitment, not satisfaction,
Additionally, one might wonder whether approach commitment emerged as a significant predictor in regression analyses with
is really something different from relationship satisfaction. How- motivational related measures. More concretely, predicting disen-
ever, there are good reasons to answer this question positively. gagement and interest for alternatives (a scale consisting of five
First, on a theoretical level, approach commitment can unambig- items, e.g., “When I meet another man/woman I consider him/her
uously be distinguished from relationship satisfaction: Satisfaction from the perspective of him/her potentially being a new partner”)
(and emotions) represent affective judgments of the relationship. resulted in rpart ⫽ .35, ␤ ⫽ ⫺.58, t(67) ⫽ ⫺2.97, p ⬍ .005, for
By contrast, approach commitment focuses on the participant’s approach commitment, and rpart ⫽ .00, ␤ ⫽ .00, t ⬍ 1, for
determination to continue the relationship. As such, it represents a satisfaction. In sum, there is a consistent and theoretically con-
motivational concept (cf. Brunstein, 1993; Fishbein & Ajzen, vincing pattern of data that speaks for the distinctiveness of ap-
1975; Gollwitzer, 1993), namely, the binding decision or intention proach commitment and satisfaction. Notwithstanding the correla-
to continue one’s intimate relationship. Second, from an empirical tion between them, each of the two constructs’ unique variance
point of view, for two concepts to be called distinct from each explains a set of different criteria variables.
other, it is crucial that the first construct still be able to explain a
substantial proportion of variance in certain third variables after
the influence of the second construct has been partialed out, 8
Some of the following variables have not been mentioned in the article
whereas this is not the case for the second construct (or at least it so far because they were not part of the central hypotheses. For the sake of
is so to a clearly lesser degree). More concretely, for our argument power and brevity, all variables used in the following regression analyses
to be valid, satisfaction— but not approach commitment—should were collapsed across the three testing periods.
218 FRANK AND BRANDSTÄTTER

Longitudinal Effects of Approach Versus Avoidance commitment at Time 1 resulted in ␤ ⫽ ⫺.15, t(50) ⫽ ⫺1.38, p ⬍
Commitment .17. The respective results for predicting avoidance commitment at
Time 2 were as follows: avoidance commitment at Time 1, ␤ ⫽
The longitudinal results of the study confirm our prediction of a .61, t(50) ⫽ 5.99, p ⬍ .001; satisfaction at Time 1, ␤ ⫽ ⫺.08, t ⬍
distinct association between approach–avoidance commitment and 1; and approach commitment at Time 1, ␤ ⫽ .20, t(50) ⫽ 1.09, p ⬎
relationship satisfaction at both Time 2 and Time 3 assessments .27. That is, in neither of the two cases did relationship satisfaction
(Hypotheses 2a). Thus, approach commitment was positively as- become a significant predictor when we controlled for the effects
sociated with relationship satisfaction 6 and even 13 months later, of approach and avoidance commitment.11
whereas avoidance commitment was negatively related with these Because we did not perform a statistical comparison of the
parameters.9 two different causality models, no conclusive statement can be
With respect to both of our emotional measures, predictions for made about the predictive value of one model in relation to the
approach versus avoidance commitment were fully confirmed for other. Instead, we argue for possible bidirectionality (see also
the Time 2 percentage of well-being and the frequency of positive Bandura’s, 1986, concept of reciprocal causality). In more
emotions, respectively (Hypotheses 2b and 2c), and we found recent publications, Rusbult and colleagues (e.g., Rusbult &
partial support in the Time 3 percentage of well-being. However, Buunk, 1993) have also started to discuss the possibility of
it remains open to speculation why neither approach nor avoidance bidirectionality. Actually, some of the variables serving to
commitment became significant predictors for the frequency of explain commitment in the theoretical formulation of the in-
positive emotions at Time 3. Notably, even in the latter analysis, vestment model (e.g., alternatives; Rusbult, 1983) sometimes
approach commitment yielded a positive beta weight as in all other also serve as variables that are contingent on commitment (e.g.,
analyses, whereas avoidance commitment carried a negative sign. derogation of alternatives; D. Johnson & Rusbult, 1989). It is
Taken together, results from these regression analyses lend interesting that, as our analyses regarding the mutual influence
support to our notion that approach commitment is positively and between partners demonstrated, there seems to be a bidirec-
avoidance commitment negatively associated with relationship tional influence between partners: One partner’s approach com-
quality.10 By using Time 1 predictors to explain the variance in mitment predicted the other partner’s relationship satisfaction,
Time 2 and Time 3 dependent measures, we were able to demon- and vice versa, this reinforcement hopefully resulting in a kind
strate that this association holds true for a time period as long as 13 of mutual escalation of satisfaction and approach commitment.
months.
These results replicate findings from studies on approach–
avoidance personal goals and subjective well-being (e.g., Elliot & Related Concepts Amenable to an Approach–Avoidance
Sheldon, 1997; Elliot et al., 1997; Roney et al., 1995) and extend Interpretation
them into the field of intimate relationships by providing empirical
Although our model is a new means of formally distinguishing
support for Rusbult’s (1991) hitherto untested assumption in the
between different types of commitment, similar concepts have
relationship domain that “people acting out of a strong positive
desire to continue a line of action should differ in serious ways
from . . . people acting out of a sense of . . . pressure” (p. 163). 9
Additional analyses were run with a composite score of only those
three items that were taken from a general commitment measure (i.e., the
three items of Brunstein, 1993, see Method section). Analogously to our
Direction of Influence
two-faceted regression approach for predicting relationship satisfaction at
In the present research we assume a direction of influence that Time 2 and Time 3, we first entered the baseline of relationship satisfaction
at Time 1, then added the newly created general commitment measure (the
goes from commitment types to relationship satisfaction. A similar
same was done for the other two dependent measures, well-being time and
view within the domain of organizational commitment and satis-
frequency of positive emotions). In all regression analyses, this general
faction is supported by research done by Vandenberg and Lance measure did not add significantly to the prediction—that is, when we
(1992). In the relationship domain, however, our position contrasts predicted relationship satisfaction at Time 2, for increment, F(1,
with that of Rusbult and colleagues (Rusbult & Buunk, 1993; 51) ⫽ 1.83, p ⬍ .19; in the remaining five analyses, all Fs for increment
Rusbult & Martz, 1995; Van Lange et al., 1997), who advocated were less than 1.
the reverse direction of influence: In their view, commitment is 10
Given the low (and nonsignificant) zero-order correlations between
determined by satisfaction with the relationship (for a summary, avoidance commitment and the respective dependent measures, one might
see Rusbult, 1991). wonder whether the effect of avoidance commitment in the regression
To test Rusbult’s approach, we assessed the predictive value of analyses reported was due to a suppressor effect of approach commitment.
the direction of influence advocated by Rusbult and colleagues However, regression analyses run with the respective baseline measure of
(Rusbult & Buunk, 1993; Rusbult & Martz, 1995; Van Lange et the dependent variable in a first step and avoidance commitment alone in
a second step revealed that avoidance commitment at Time 1 still added to
al., 1997). Analogously to testing our model, we computed regres-
the prediction of Time 2 satisfaction, ␤ ⫽ ⫺.20, F(54) ⫽ 2.7, p ⫽ .10;
sion analyses predicting approach and avoidance commitment,
well-being time, ␤ ⫽ ⫺.31, F(54) ⫽ 6.6, p ⫽ .01; and frequency of
respectively, at Time n by entering approach and avoidance com- emotions, ␤ ⫽ ⫺.20, F(54) ⫽ 3.1, p ⫽ .08, respectively.
mitment at Time 1 as well as satisfaction with the relationship at 11
Conceptually identical results were obtained for predicting Time 3
Time 1 (see lower part of Figure 1 for a path model presentation). approach and avoidance commitment on the basis of Time 1 relationship
Predicting Time 2 approach commitment, approach commitment at satisfaction and approach or avoidance commitment, respectively. Further-
Time 1 resulted in ␤ ⫽ .86, t(50) ⫽ 4.48, p ⬍ .001, satisfaction at more, results using well-being time or frequency of emotions as predictors
Time 1 resulted in ␤ ⫽ .23, t(50) ⫽ 1.20, p ⬍ .24, and avoidance for approach and avoidance commitment parallel those reported above.
 APPROACH VERSUS AVOIDANCE COMMITMENT 219

been introduced in the past by other social scientists. For example, commitment negatively associated with relationship quality—
Levinger (1979) discussed a notion of attractions versus barriers would also be found in non-Western cultures. “The hedonic prin-
that can be interpreted in terms of approach versus avoidance ciple that people approach pleasure and avoid pain has been the
commitment. Despite the conceptual overlap between Levinger’s basic motivational principle throughout the history of psychology,
model and ours, there is one important difference between them. with ancient roots” (Higgins, 1998, p. 1). Moreover, approach and
As documented in Levinger’s (1979) comprehensive review of the avoidance with their specific affective concomitants can even be
literature on marital cohesiveness and divorce, attractions and distinguished on a neurobiological level (e.g., Gray, 1982). There-
barriers both promote marital stability indiscriminately. Hence, fore, the distinction between approach and avoidance as basic
Levingers’s conceptual distinction between attractions and barriers motivational tendencies should be valid among all human beings.
seems less illuminating than our explication of the distinct effects What might differ between Western independent and non-Western
of different types of commitment on relationship satisfaction and interdependent cultures (Markus & Kitayama, 1991), though, is the
emotionality, which is new to our theorizing. Simpson (1990) also content of approach- versus avoidance-motivated commitment as
discussed a notion similar to our approach–avoidance distinction well as their relative weight in predicting outcome variables. For
in his work. However, his secure versus avoidant attachment styles example, interdependence, care for others, and harmony might be
in romantic associations are conceptualized as a relatively stable more important and positively regarded values in Eastern cultures,
individual-differences variable originating in early socialization. whereas they might disappear behind more individualistic issues in
The specific and possibly changing nature of commitment a person Western, independent cultures. Yet the question of cultural gener-
might experience in an ongoing relationship over the course of alizability remains an interesting issue for future research.
time or in different intimate relationships cannot be captured by Fifth, the present study tells us little about how approach–
Simpson’s concept. avoidance commitment is related to partnership stability. Only 3
couples who answered the second and/or third questionnaire indi-
Limitations and Directions for Future Research cated that they had separated. Therefore, we can only speculate
about the distinct effects of approach versus avoidance commit-
Clearly, there are some limitations and unresolved issues inher- ment on the continuity of relationships on the basis of prior
ent in the present research, some of them indicating avenues for research and theorizing. For example, Levinger (1979) pointed out
future research. First, our analysis does not systematically address that “barriers are important to keep long-term relationships in-
the question of which factors contribute to the development of tact. . . . Barriers lessen the effect of temporary fluctuations in
approach and avoidance commitment. The pattern of correlation interpersonal attraction; even if attraction becomes negative, bar-
with values, relationship duration, and similarity between partners’ riers act to continue the relationship” (p. 41). This should be a
conceptions of a good relationship hints at a combined influence of caveat for the reader not to conclude immediately from our results
more stable individual-differences variables and relationship- on relationship quality measures that avoidance commitment is
specific situational determinants. exclusively negative in its effects. Instead, this type of commit-
As a second unresolved issue, one might ask why avoidance ment might serve an important and beneficial function in relation-
commitment is negatively related to relationship satisfaction. ship stability.
Drawing on the finding that avoidance motivation leads to a
heightened sensitivity for negative information (Higgins, 1998; Summary and Conclusions
Higgins & Tykocinski, 1992), one might speculate that avoidance-
driven commitment makes a partner’s negative attributes and In conclusion, the purpose of the present research is to provide
actions or the incurred cost of the relationship more salient. Such evidence for an approach versus avoidance type of commitment
focusing on cost is known to be a characteristic of unsatisfactory and its link to relationship quality parameters. The study yields
relationships (Gottman, 1998). Further research would have to encouraging results, suggesting that approach and avoidance com-
analyze such mediating processes on a more microscopic level. mitment are opposite and powerful predictors explaining variance
Third, related to the previous issue, experimental laboratory of relationship quality parameters.
research is needed to circumvent the methodological limitation of In sum, we believe that applying the approach–avoidance per-
our study. Research on close relationships might substantially spective in the field of commitment to personal relationships has
benefit from multimethod approaches, including observational and several theoretical advantages. It is a new approach to the study of
experimental methods beyond simple self-report measures (Clark commitment and satisfaction in close relationships, although our
& Reis, 1988; see also Berscheid, 1994). Additionally, the use of model’s basic concepts are firmly rooted in existing psychological
diary studies might be a promising avenue of exploration to literature. It is parsimonious, as several theoretical accounts of
illuminate how approach and avoidance commitment affect other commitment and related concepts (M. Johnson, 1991; Levinger,
important relationship variables in everyday relationship 1979; Meyer & Allen, 1991; Simpson, 1990) can be integrated
functioning. under a single theoretical construct. Moreover, it incorporates
Fourth, the interpretation of our results is restricted to samples social–psychological theorizing on relationship functioning with
that are similar in age, education level, and ethnical– cultural motivational constructs of goal striving. An even more important
background. As Markus and Kitayama (1991) have argued in their advantage is that drawing on the approach–avoidance distinction
programmatical review, many social–psychological phenomena provides a rich understanding of basic motivational processes,
once thought to be universally valid only hold true in Western which are of crucial importance not only in self-regulation but also
cultures. However, we still argue that our basic finding—accord- with respect to the functioning of intimate relationships. Approach
ing to which approach commitment is positively and avoidance and avoidance motivation affect emotional regulation, which, in
220 FRANK AND BRANDSTÄTTER

turn, determines relationship quality, which is a core determinant personality–well-being relationship. Journal of Personality and Social
of global life satisfaction as well as of mental and physical health Psychology, 75, 1282– 1299.
(Christensen & Heavey, 1999; Gottman, 1998). We therefore Elliot, A. J., Sheldon, K. M., & Church, M. (1997). Avoidance personal
strongly argue that researchers in the field of commitment should goals and subjective well-being. Personality and Social Psychology
Bulletin, 23, 915–927.
consider these different commitment bases in their research.
Emmons, R. A., & Kaiser, H. A. (1996). Goal orientation and emotional
well-being: Linking goals and affect through the self. In L. L. Martin &
References A. Tesser (Eds.), Striving and feeling: Interactions among goals, affect,
and self-regulation (pp. 79 –98). Mahwah, NJ: Erlbaum.
Antill, J. K. (1983). Sex role complementarity versus similarity in married Feather, N. T. (1990). Bridging the gap between values and actions. In
couples. Journal of Personality and Social Psychology, 45, 145–155. E. T. Higgins & R. M. Sorrentino (Eds.), Handbook of motivation and
Arkes, H. R., & Blumer, C. (1985). The psychology of sunk cost. Orga- cognition (Vol. 2, pp. 151–192). New York: Guilford Press.
nizational Behavior and Human Decision Processes, 35, 124 –140. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior:
Atkinson, J. W. (1957). Motivational determinants of risk-taking behavior. An introduction to theory and research. Reading, MA: Addison-Wesley.
Psychological Review, 63, 359 –372. Glenn, N. D. (1990). Quantitative research on marital quality in the 1980s:
Bandura, A. (1986). Social foundations of thought and action: A social A critical review. Journal of Marriage and the Family, 52, 818 – 831.
cognitive theory. Englewood Cliffs, NJ: Prentice Hall. Gollwitzer, P. M. (1993). Goal achievement: The role of intentions. In W.
Becker, T. E. (1992). Foci and bases of commitment: Are they distinctions Stroebe & M. Hewstone (Eds.), European review of social psychology
worth making? Academy of Management Journal, 35, 232–244. (Vol. 4, pp. 141–185). Chichester, England: Wiley.
Berscheid, E. (1994). Interpersonal relationships. Annual Review of Psy- Gonzalez, R., & Griffin, D. (1997). On the statistics of interdependence:
chology, 45, 79 –129. Treating dyadic data with respect. In S. Duck (Ed.), Handbook of
Brickman, P. (1987). Commitment. In C. B. Wortman & R. Sorrentino personal relationships (pp. 271–302). Chichester, England: Wiley.
(Eds.), Commitment, conflict, and caring (pp. 1–18). Englewood Cliffs, Gottman, J. M. (1998). Psychology and the study of marital processes.
NJ: Prentice-Hall. Annual Review of Psychology, 49, 169 –197.
Brockner, J. (1992). The escalation of commitment to a failing course of Gray, J. A. (1982). The neuropsychology of anxiety: An enquiry into the
action: Toward theoretical progress. Academy of Management Re- functions of the septohippocampal system. New York: Oxford University
view, 17, 39 – 61. Press.
Brunstein, J. C. (1993). Personal goals and subjective well-being: A Griffin, D., Murray, S., & Gonzalez, R. (1999). Difference score correla-
longitudinal study. Journal of Personality and Social Psychology, 65, tions in relationship research: A conceptual primer. Personal Relation-
1061–1070. ships, 6, 505–518.
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear models: Hassebrauck, M. (1991). ZIP— ein Instrumentarium zur Erfassung der
Applications and data analysis methods. Newbury Park, CA: Sage. Zufriedenheit in Paarbeziehungen [ZIP—A scale for assessment of
Bui, K., Peplau, L. A., & Hill, C. T. (1996). Testing the Rusbult model of satisfaction in close relationships]. Zeitschrift für Sozialpsychologie, 22,
relationship commitment and stability in a 15-year study of heterosexual 256 –259.
couples. Personality and Social Psychology Bulletin, 22, 1244 –1257. Hassebrauck, M. (1996). Beziehungskonzepte und Beziehungszufrieden-
Byrne, D. (1971). The attraction paradigm. New York: Academic Press. heit: Die Bedeutung tatsächlicher und wahrgenommener Konzeptähn-
Byrne, D., Clore, G. L., & Smeaton, G. (1986). The attraction hypothesis: lichkeit in Paarbeziehungen [Relationship among concepts and relation-
Do similar attitudes affect anything? Journal of Personality and Social ship satisfaction: The importance of real and perceived similarity of
Psychology, 51, 1167–1170. couples]. Zeitschrift für Sozialpsychologie, 27, 183–192.
Carver, C. S., & Scheier, M. F. (1990). Principles of self-regulation: Action Heckhausen, H. (1991). Motivation and action. Berlin: Springer.
and emotion. In E. T. Higgins & R. M. Sorrentino (Eds.), Handbook of Hendrick, S. S. (1988). A generic measure of relationship satisfaction.
motivation and cognition (Vol. 2, pp. 3–52). New York: Guilford Press. Journal of Marriage and the Family, 50, 93–98.
Christensen, A., & Heavey, C. L. (1999). Interventions for couples. Annual Hendrick, S. S., & Hendrick, C. (1992). Romantic love. Thousand Oaks,
Review of Psychology, 50, 165–190. CA: Sage.
Clark, M. S., & Reis, H. T. (1988). Interpersonal processes in close Higgins, E. T. (1998). Promotion and prevention: Regulatory focus as a
relationships. Annual Review of Psychology, 39, 609 – 672. motivational principle. In M. P. Zanna (Ed.), Advances in experimental
Cohen, J. P., & Cohen, P. (1983). Applied multiple regression/correlation social psychology (Vol. 30, pp. 1– 46). San Diego, CA: Academic Press.
analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum. Higgins, E. T., Roney, C. J. R., Crowe, E., & Hymes, C. (1994). Ideal vs.
Drigotas, S. M., & Rusbult, C. E. (1992). Should I stay or should I go? A ought predilections for approach and avoidance: Distinct self-regulatory
dependence model of breakups. Journal of Personality and Social Psy- systems. Journal of Personality and Social Psychology, 66, 276 –286.
chology, 62, 62– 87. Higgins, E. T., & Tykocinski, O. (1992). Self-discrepancies and biograph-
Drigotas, S. M., Rusbult, C. E., & Verette, J. (1999). Level of commitment, ical memory: Personality and cognition at the level of psychological
mutuality of commitment, and couple well-being. Personal Relation- situation. Personality and Social Psychology Bulletin, 18, 527–535.
ships, 6, 389 – 409. Huston, T. L., & Levinger, G. (1978). Interpersonal attraction and rela-
Elliot, A. J., & Church, M. (1997). A hierarchical model of approach and tionships. Annual Review of Psychology, 29, 115–156.
avoidance achievement motivation. Journal of Personality and Social Johnson, D. J., & Rusbult, C. E. (1989). Resisting temptation: Devaluation
Psychology, 72, 218 –232. of alternative partners as a means of maintaining commitment in close
Elliot, A. J., & Harackiewicz, J. M. (1996). Goal setting, achievement relationships. Journal of Personality and Social Psychology, 57, 967–
orientation, and intrinsic motivation: A mediational analysis. Journal of 980.
Personality and Social Psychology, 66, 968 –980. Johnson, M. (1991). Commitment to personal relationships. Advances in
Elliot, A. J., & Sheldon, K. M. (1997). Avoidance achievement motivation: Personal Relationships, 3, 117–143.
A personal goals analysis. Journal of Personality and Social Psychol- Judd, C. M., McClelland, G. H., & Culhane, S. E. (1995). Data analysis:
ogy, 73, 171–185. Continuing issues in the everyday analysis of psychological data. Annual
Elliot, A. J., & Sheldon, K. M. (1998). Avoidance personal goals and the Review of Psychology, 46, 433– 465.
 APPROACH VERSUS AVOIDANCE COMMITMENT 221

Kenny, D. A. (1996). Models of non-independence in dyadic research. Rusbult, C. E. (1991). Commentary on Johnson’s ‘commitment to personal
Journal of Social and Personal Relationships, 13, 279 –294. relationships’: What’s interesting, and what’s new? Advances in Per-
Levinger, G. (1979). A social psychological perspective on marital disso- sonal Relationships, 3, 151–169.
lution. In G. Levinger & O. C. Moles (Eds.), Divorce and separation. Rusbult, C. E., & Buunk, B. P. (1993). Commitment processes in close
Context, causes, and consequences (pp. 37– 60). New York: Basic relationships: An interdependence analysis. Journal of Social and Per-
Books. sonal Relationships, 10, 175–204.
Lund, M. (1985). The development of investment and commitment scales Rusbult, C. E., & Martz, J. M. (1995). Remaining in an abusive relation-
for predicting continuity of personal relationships. Journal of Social and ship: An investment model analysis of nonvoluntary dependence. Per-
Personal Relationships, 2, 3–23. sonality and Social Psychology Bulletin, 21, 558 –571.
Lydon, J. E. (1996). Toward a theory of commitment. In C. Seligman, J. Rusbult, C. E., Verette, J., Whitney, G. A., Slovik, L. F., & Lipkus, I.
Olson, & M. Zanna (Eds.), Values: The Eighth Ontario Symposium (pp. (1991). Accommodation processes in close relationships: Theory and
191–213). Hillsdale, NJ: Erlbaum. preliminary evidence. Journal of Personality and Social Psychology, 60,
Lydon, J. E., Meana, M., Sepinwall, D., Richards, N., & Mayman, S. 53–78.
(1999). The commitment calibration hypothesis: When do people de- Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the
value attractive alternatives? Personality and Social Psychology Bulle- facilitation of intrinsic motivation, social development, and well-being.
tin, 25, 152–161. American Psychologist, 55, 68 –78.
Lydon, J. E., Pierce, T., & O’Regan, S. (1997). Coping with moral Schmalt, H. -D. (1999). Assessing the achievement motive using the grid
commitment to long-distance relationships. Journal of Personality and technique. Journal of Research in Personality, 33, 109 –130.
Social Psychology, 73, 104 –113. Schwartz, S. H. (1995). Values. In A. S. R. Manstead & M. Hewstone
Markus, H., & Kitayama, S. (1991). Culture and the self: Implications for (Eds.), The Blackwell encyclopedia of social psychology (pp. 665– 667).
cognition, emotion, and motivation. Psychological Review, 98, 224 – Oxford, England: Blackwell.
253. Schwartz, S. H., & Bilsky, W. (1990). Toward a theory of the universal
Matthieu, J. E., & Zajac, D. M. (1990). A review and meta-analysis of the content and structure of values: Extensions and cross-cultural replica-
antecedents, correlates, and consequences of organizational commit- tions. Journal of Personality and Social Psychology, 58, 878 – 891.
ment. Psychological Bulletin, 108, 171–194. Schwartz, S. H., & Sagiv, L. (1995). Identifying culture-specifics in the
Meyer, J. P., & Allen, N. J. (1991). A three-component conceptualization content and structure of values. Journal of Cross-Cultural Psychology,
of organizational commitment. Human Resource Management Re- 26, 92–116.
view, 1, 61–98. Simpson, J. A. (1990). Influence of attachment style on romantic relation-
Meyer, J. P., Allen, N. J., & Smith, C. A. (1993). Commitment to orga- ships. Journal of Personality and Social Psychology, 59, 971–980.
nizations and occupations: Extension and test of a three-component Simpson, J. A., Gangestad, S. W., & Lerma, M. (1990). Perception of
conceptualization. Journal of Applied Psychology, 78, 538 –551. physical attractiveness: Mechanisms involved in the maintenance of
Miller, S. R. (1997). Inattentive and contended: Relationship commitment romantic associations. Journal of Personality and Social Psychol-
and attention to alternatives. Journal of Personality and Social Psychol- ogy, 59, 1192–1201.
ogy, 73, 758 –766. Somers, M. J. (1995). Organizational commitment, turnover and absentee-
Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996). The benefits of ism: An examination of direct and interaction effects. Journal of Orga-
positive illusions: Idealization and the construction of satisfaction in nizational Behavior, 16, 49 –58.
close relationships. Journal of Personality and Social Psychology, 70, Sternberg, R. J. (1986). A triangular theory of love. Journal of Personality
79 –98. and Social Psychology, 93, 119 –135.
Novacek, J., & Lazarus, R. S. (1990). The structure of personal commit- Vandenberg, R. J., & Lance, C. E. (1992). Examining the causal order of
ments. Journal of Personality, 58, 693–715. job satisfaction and organizational commitment. Journal of Manage-
Roney, C. J. R., Higgins, E. T., & Shah, J. (1995). Goals and framing: How ment, 18, 153–167.
outcome focus influences motivation and emotion. Personality and Van Lange, P. A. M., Rusbult, C. E., Drigotas, S. M., Arriaga, X. B.,
Social Psychology Bulletin, 21, 1151–1160. Witcher, B. S., & Cox, C. L. (1997). Willingness to sacrifice in close
Rosenbaum, M. E. (1986). The repulsion hypothesis: On the nondevelop- relationships. Journal of Personality and Social Psychology, 72, 1373–
ment of relationships. Journal of Personality and Social Psychology, 51, 1395.
1156 –1166. Wieselquist, J., Rusbult, C. E., Foster, C. A., & Agnew, C. R. (1999).
Rusbult, C. E. (1980). Commitment and satisfaction in romantic associa- Commitment, pro-relationship behavior, and trust in close relationships.
tions: A test of the investment model. Journal of Experimental Social Journal of Personality and Social Psychology, 77, 942–966.
Psychology, 16, 172–186.
Rusbult, C. E. (1983). A longitudinal test of the investment model: The
development (and deterioration) of satisfaction and commitment in Received April 24, 2000
heterosexual involvements. Journal of Personality and Social Psychol- Revision received June 8, 2001
ogy, 45, 101–117. Accepted July 31, 2001 䡲