.

i fllMMlD~GlOUP

Resisting Temptation: Devaluation of Alternative Partners as a Means

of Maintaining Commitment in Close Relationships
Dennis J. Johnson Caryl E. Rusbult
Microsoft Corporation, Redmond, Washington University of North Carolina at Chapel Hill

This work tested the hypothesis that persons who are more committed to their relationships devalue
potential alternative partners, especially attractive and threatening alternatives. In Study 1, a longitu-
dinal study, perceived quality of alternatives decreased over time among stayers but increased for
leavers. In Study 2, a computer dating service paradigm, more committed persons exhibited greatest
devaluation of alternatives under conditions of high threat—when personally evaluating extremely
attractive alternative partners. In Study 3, a simulation experiment, the tendency to reject and de-
value alternatives was greater under conditions of high commitment. In all three studies, tendencies
to devalue were more strongly linked to commitment than to satisfaction.

Virtue consists, not in abstaining from vice, but in not desiring it. enjoyment of an alternative relationship, even if brief and su-
—George Bernard Shaw, perficial, may hurt their partners or harm their relationships—
Maxims for Revolutionists
and they wish to avoid such damage—their options include the
P and O meet, fall in love, become committed, and live hap- following: P can invest numerous irretrievable resources in the
pily ever after. Is that the end of the story? Probably not. First, relationship with O, so as to create barriers against the alterna-
it is likely that P's and O's feelings of commitment to their rela- tive. P and O can pledge their mutual fidelity—perhaps even
tionship will vary over time, sometimes falling to "threaten- publicly—and rely on virtue to assist them in weathering the
able" low levels. Second, it is likely that P and O will occasion- storm. O can adopt a broad time perspective, recognizing that
ally encounter attractive alternative partners who challenge although enjoyment of the alternative might bring short-term
their commitment. Thus, it is important to gain a greater under- pleasure, long-term gratification requires fidelity. These and
standing of the processes by which individuals attempt to main- other mechanisms may be helpful in the struggle with tempta-
tain committed relationships. Given that most extant theories tion. The current research concerns yet another means of deal-
of developing relationships treat the phenomenon of commit- ing with temptation and maintaining commitment to ongoing
ment as an outcome (Johnson, 1973; Lund, 1985; Rusbult, relationships: devaluation of attractive alternative partners.
1980)—as a dependent variable of interest in and of itself— Social scientists have long recognized that one of the major
little is known about the dynamics by which persons protect threats to the stability of a relationship is the presence of an
and maintain their commitments. As Kelley (1983) noted, it is attractive alternative (Kelley, 1983;Leik&Leik, 1977; Rusbult,
important that the "processes that promote the person's being 1983; Thibaut & Kelley, 1959). Indeed, research on commit-
in a state in which the causal conditions favoring continued ment in close relationships has demonstrated that persons who
membership stably outweigh those acting against it" (p. 296) be believe they have attractive alternatives report lower commit-
explored. ment to maintain their current relationships; individuals with
When highly committed persons are confronted with attrac- less attractive alternatives are more likely to retain strong com-
tive alternative partners, what are their options? By what pro- mitment (Rusbult, 1980). More important, it has been demon-
cesses do individuals manage to maintain commitment? Of strated that perceived quality of alternatives decreases consider-
course, one option is to surrender to temptation, enjoy the alter- ably with increased involvement: In a longitudinal study of dat-
native, and suffer the damage. For some relationships, the dam- ing relationships, Rusbult (1983) found that as individuals
age may be slight. However, when individuals believe that the became increasingly involved with their partners, they de-
scribed their alternatives in increasingly negative terms. These
findings suggest that changes in evaluations of alternative part-
We wish to express our gratitude to Elizabeth Farmer, Lisa Gibbs, ners may play a role in maintaining commitment to current
Benton Seay, and Karin Shea for their help in conducting Study 2 and partners. Through what process do such changes come about?
to Harry Reis and several anonymous reviewers for their exceptionally
helpful comments on an earlier version of this article. Correspondence
Rusbult (1983) argued that the tendency of committed per-
concerning this article should be addressed to Caryl E. Rusbult, Depart- sons to report decreased attraction to alternatives may be due
ment of Psychology, University of North Carolina, Chapel Hill, North to two processes: First, it may be that alternatives really do de-
Carolina 27599-3270. cline over time; alternative partners may be reluctant to ap-
Journal of Personality and Social Psychology, 1989, Vol. 57, No. 6,967-980
Copyright 1989 by the American Psychological Association, Inc. 0022-3J 14/89/S00.75

967
968 DENNIS J. JOHNSON AND CARYL E RUSBULT

proach a committed individual because of awareness of his or nation thus argues that the tendency to devalue alternatives
her involvement. As Kelley (1983) noted, "other persons who emerges from the experience of happiness and satisfaction in an
might have been available as partners now take themselves out ongoing relationship.
of the running and look elsewhere for associations" (p. 305). A Thus, two lines of reasoning—one motivational and the other
second, cognitive explanation argues for changes in individuals' perceptual—support the assertion that more highly committed
thoughts regarding available alternatives. Many theorists have persons may devalue alternative partners. The comparison-level
argued that committed persons may maintain stable relation- account implies more of a perceptual phenomenon, whereby
ships by devaluing alternative partners (Ranter, 1968; Kelley, alternatives "look less good" to the highly committed because
1983; Leik & Leik, 1977; Thibaut & Kelley, 1959). As Thibaut their expectations have been inflated as a consequence of in-
and Kelley (1959) proposed, "[conflict] can be reduced by di- volvement in a very satisfying relationship. The motivational
minishing the value of the unattainable [person]. . .bytaking explanation (e.g., a dissonance account) implies that the pres-
a 'sour grapes' attitude toward the rewarding aspects of the in- ence of an attractive alternative produces conflict for the com-
teraction or by emphasizing the negative, cost-increasing as- mitted individual, and that this conflict may be reduced or elim-
pects of it" (p. 175). If alternative partners merely look less ap- inated by devaluing that alternative. One might further reason
pealing to the committed individual, the relationship is pro- that tendencies to devalue alternatives should be greatest when
tected. In indirect support of this assertion, Ranter (1968) the alternative produces greater conflict.
found that in Utopian communities, outsiders are described in Conflict should be increased by at least two factors. First,
negative terms and alternatives are renunciated. Such processes more tempting alternatives should induce greater conflict In-
presumably ensure continued commitment to the community. deed, research on postdecisional spreading of alternatives dem-
Unfortunately, devaluation of alternatives in romantic relation- onstrates that when an alternative is very attractive in compari-
ships has not heretofore been examined. son to the chosen option, tendencies to disparage the nonchosen
At least two lines of reasoning would lead one to predict that alternative are greater (Brehm, 1956). Second, more available
highly committed persons may devalue alternatives. First, a mo- alternatives should induce greater tension. Devaluation of alter-
tivational explanation—such as that proffered by dissonance natives by committed persons should be most pronounced
theory—would suggest that when important beliefs are in con- when the alternative is a realistic threat. Although one may ac-
flict, changes in cognitions may occur toward the goal of restor- knowledge that a remote and unattainable alternative is attrac-
ing consonance (Festinger, 1957; Greenwald & Ronis, 1978). tive, no tension is experienced because that person is no real
The committed person's belief that "I am a loyal and commit- threat to one's current relationship. It is when one is faced with
ted partner" would be dissonant with the belief "I am attracted an actual choice between a partner and an alternative—when
to an alternative partner." One means of decreasing such con- one must actively consider the possibility of forming a relation-
flict is to reduce the perceived attractiveness of the alternative, ship with another—that threat should be greatest and devalua-
by disparaging some personal quality of the alternative or by tion of alternatives should be most pronounced.
devaluing something about the relationship that might be estab- Thus, we predict that greater commitment is associated with
lished with that person. The cognition "That person would tendencies to devalue alternative partners, and that this process
probably cheat on me and make me blue" is not dissonant with is most marked when the alternative poses the greatest threat to
a belief that one is a loyal and committed partner. For less the current relationship: when the alternative is exceptionally
attractive, and when the individual is faced with an actual op-
committed persons, attraction to an alternative produces little
portunity to become involved with that person. We also attempt
conflict, so devaluation of that person is unnecessary. Thus, mo-
to determine whether this process is more closely tied to feelings
tivational explanations argue that the tendency to devalue alter-
of satisfaction—a finding that would imply a simple perceptual
natives emerges from the desire to protect an ongoing com-
process whereby devaluation results from reduced comparison
mitment.
level—or whether this process is more closely tied to feelings of
A second account of this process is more of a perceptual ex-
commitment—a finding that would imply a motivated process
planation, and it rests on the concept of comparison level, or
based more on the experience of threat to an ongoing commit-
expectations regarding the quality of close relationships (Thi-
ment.
baut & Kelley, 1959). Persons with high comparison levels
Three complementary studies explore this phenomenon.
should view alternatives as less appealing than should persons
Study 1 uses data from a 7-month longitudinal study to exam-
with low expectations. Indeed, Kenrick and Gutierres (1980)
ine reactions to alternatives as they exist in everyday life. Study
found that men who were exposed to extremely attractive
2 examines the impact of variations in commitment and satis-
women later judged potential blind dates as less attractive than
faction on evaluations of potential dating partners who vary in
did a comparable control group. Given that committed persons
attractiveness and degree of threat. Study 3 actively manipu-
are often very satisfied with their relationships, and given that
lates commitment and satisfaction in a simulation experiment
persons may use their current relationships as a standard for
and obtains evaluations of attractive and threatening alterna-
comparison, more committed individuals should be more likely
tives. Together, the three studies maximize both internal and
to judge that alternatives fall short of expectations. Indeed, Thi-
external validity and should provide a good test of our hypoth-
baut and Kelley (1959) argued that comparison level is "condi-
esis.
tioned partly by outcomes in the present relationship" (p. 82)
and that "the more satisfactory any given relationship has been Study 1
found to be, the higher will be the comparison level for evaluat- Study 1 is a reanalysis of data obtained in Rusbult's (1983)
ing any new relationship" (p. 95). The comparison-level expla- 7-month longitudinal study of college-age dating relationships.
 RESISTING TEMPTATION 969

We examined several measures of perceived quality of alterna- The commitment measures were "How likely is it that you will end your
tives, the goal being to determine the strength of the link be- relationship in the near future?" (1 = not at all likely, 9 = extremely
tween increasing commitment to current partners and declin- likely; reversed), "For what length of time would you like your relation-
ship to last?" (1 = week or so, 9 = lifetime), "How attractive an alterna-
ing evaluations of alternatives, especially tendencies to devalue
tive would you require before adopting it and ending your relationship?"
actual alternative partners. We also examine the impact of satis-
(1 = moderately attractive alternative, 9 = extremely attractive alterna-
faction level on evaluations of alternatives.
tive), "To what extent are you 'attached' to your partner?" (1 = not at
all, 9 = extremely), and "To what extent are you committed to your
Method relationship?" (1 = not at all, 9 = extremely).
Reliability of measures. We computed reliability coefficients for the
Subjects and procedure. Subjects were 17 men and 17 women at set of items designed to measure each construct These analyses re-
Franklin and Marshall College who volunteered to participate in re- vealed sizable alphas for the global measures of satisfaction (.89), com-
sponse to flyers placed in their campus mailboxes. Out of 119 volun- mitment (.90), and alternative quality (.84) and for the pairs of concrete
teers, actual subjects were selected on the basis of sex (half were men measures designed to measure evaluations of actual alternative partners
and half were women) and initial duration of the relationship (5 men (.84) and evaluations of spending time alone (.81). Therefore, a single
and 5 women in the 0-2 week category; 4 men and 4 women in the 2- averaged measure of each construct was formed.
4, 4-6, and 6-8 week categories). Only one partner in a relationship
was allowed to participate. Subjects' mean age was 19.79, and the mean
Results
duration of their relationships at the start of the study was 4.15 weeks.
Subjects were paid $2.50 for each questionnaire they completed. Changes over time in evaluations of alternatives. We per-
The study began near the start of the 1977-1978 academic year, and formed regression analyses that included the two measures of
participation concluded when the relationship ended or the study itself
evaluations of alternatives as dependent variables and time as
ended. One man dropped out of the study at Time 2, and 2 men and 1
an independent variable. To control for the nonindependence of
woman stopped responding at Time 8 (spring break). Of the remaining
multiple measures obtained over time from a given individual,
30 subjects, 10 persons' relationships ended (4 men and 6 women), and
20 subjects responded throughout (10 men and 10 women). Subjects subject number was included as a categorical variable (Cohen
returned questionnaires through campus mail within 4 days of their & Cohen, 1975). As reported in Rusbult (1983), these analyses
receipt. If the subject did not return a questionnaire within 7 days of revealed that perceived quality of alternatives declined signifi-
the time it was mailed, he or she was sent a reminder. This procedure cantly over time (ft = -0.179, p < .002). Does this effect result
was altered only twice: during winter break (1978), when questionnaires from a tendency to evaluate alternative partners more nega-
with stamped return envelopes were mailed to subjects' off-campus ad- tively as a result of increasing commitment? We reasoned that
dresses, and during the "radiation vacation" occasioned by the 1979 if the current model has merit, the decline in perceived quality
Three Mile Island incident, when questionnaires were delayed 4 days
of alternatives should occur for perceptions of potential alterna-
because the college was closed.
tive partners, but not necessarily for feelings regarding spending
Questionnaires. The questionnaires were modeled after the instru-
time alone (i.e., noninvolvement). That is, an actual "in the
ment used by Rusbult (1980). For each model variable, subjects first
completed items that represented concrete operationauzations of the flesh" challenger is presumably more threatening than the op-
variable and then answered global measures tapping that variable. The tion of solitude. Furthermore, the decline in evaluations of al-
concrete measures prepared subjects to answer the global questions and ternative partners should be stronger for persons who remained
"taught" them the meaning of the global items. Complete information committed to their relationships throughout the study than for
regarding the questionnaire is presented in Rusbult (1983). persons who ended their relationships; those who remained
Four concrete and two global items measured alternative quality. Two committed should be more likely to have engaged in devalua-
concrete items obtained evaluations of actual alternative partners: "In tion of alternatives.
terms of the sorts of rewards and costs discussed above (e.g.. intelli-
Following these lines of reasoning, we regressed evaluations
gence, sense of humor, physical attractiveness), how appealing are the
of actual alternative partners and evaluations of spending time
people other than your partner with whom you could become in-
alone onto time, including subject number as a categorical vari-
volved?" (1 = very, 9 = not at all; revased) and "In terms of the sorts
of rewards and costs discussed above, how difficult would it be to replace able. To determine whether changes over time differed for "stay-
your partner?" (1 = impossible, 9 = not at all difficult). Two concrete ers" and "leavers," the regression models also included terms
items obtained evaluations of the alternative of spending time without for stayer-versus-leaver and the Stayer-Leaver X Time interac-
a romantic partner: "How important is it to you to be involved in a tion (0 = stayers, 1 = leavers). The results of these analyses are
relationship?" (1 = not at all, 9 = extremely; reversed) and "To what summarized in Table 1. Consistent with the logic outlined ear-
extent can you be happy when you are not involved in a romantic rela- lier, the analyses revealed that in predicting evaluations of alter-
tionship?" (1 = 7 can be very happy, 9 = / am extremely unhappy; re- native partners, the unstandardized regression coefficient for
versed). The global measures were "In general, how appealing are your time was significantly negative (B = -0.125, p < .001). Further-
alternatives (dating another person or other persons or being without
more, the unstandardized regression coefficient for the Stayer-
a romantic involvement)?" (1 = not at all appealing, 9 = extremely
Leaver x Time interaction was significantly positive (B =
appealing) and "All things considered, how do your alternatives com-
0.449, p < .001); this decline was significantly less acute for leav-
pare to your current relationship?" (1 = this is much better, 9 = alterna-
tives are much better). ers. ' Indeed, as can be seen in Figure 1, whereas stayers' evalua-
Only global measures of satisfaction and commitment were obtained.
The satisfaction measures were "How much do you like your partner?"
1
(1 = not at all, 9 = very much), "To what extent are you attracted to Further analyses were performed that included main effects and in-
your partner?" (1 = not at all, 9 = extremely), and "To what degree are teractions involving gender (0 = women, 1 = men). These analyses re-
you satisfied with your relationship?" (1 = not at all, 9 = extremely). vealed that although both male and female stayers' evaluations of alter-
970 DENNIS J. JOHNSON AND CARYL E. RUSBULT

Table 1 Commitment, satisfaction, and evaluations of alternatives.

Evaluations of Actual Alternative Partners as a Function of Are evaluations of alternative partners directly linked to feel-
Time, Commitment Level, ami Satisfaction Level: Study 1 ings of commitment to current relationships? We examined the
relation between commitment and evaluations of alternatives
Multiple regression
by regressing evaluations of actual alternative partners and eval-
analyses B R F df p<
uations of spending time alone onto commitment, including
Changes over time in subject number as a categorical variable.2 Consistent with pre-
evaluations of dictions, the link with commitment was significant for evalua-
alternative partners tions of alternative partners ( B = -0.462, p < .001), but not for
Time -0.125'* .885 29.91 32,263 .001
spending time alone (B = -0.027, p < .321).3 To determine
Stayer- Leaver 0.000
Time X Stayer-Leaver 0.449" whether commitment or satisfaction more directly mediates
tendencies to devalue alternatives, we performed further analy-
Commitment and
ses that included the satisfaction variable. As for commitment,
satisfaction with
alternative partners the link with satisfaction was significant for evaluations of alter-
Commitment level -0.185" .924 48.64 35,291 .001 native partners (B = -0.242, p < .001), but not for spending
Satisfaction level -0.176" time alone (B = -0.047, p < .301).
Commitment level
As the satisfaction and commitment measures were moder-
only -0.462" .921 48.22 34,292 .001
Satisfaction level only -0.242" .910 41.31 34,292 .001 ately collinear, we performed simultaneous regressions to deter-
mine which construct more powerfully mediates devaluation,
Note. Changes over time in evaluations of alternative partners rows re- including both predictor variables. The model that included
port the results of an analysis using evaluations of alternative partners both commitment and satisfaction did not significantly predict
as the criterion and using time, stayer versus leaver (stayer = 0, leaver =
subjects' evaluations of spending time alone (commitment B =
1), and the Time X Stayer-Leaver interaction as predictors. Commit-
ment and satisfaction with alternative partners rows report the results -0.023, p < .431; satisfaction B = -0.011, p < .857), but both
of three analyses using evaluations of alternative partners as the crite- variables contributed significantly to predicting evaluations of
rion: one with commitment as the predictor, one with satisfaction as the actual alternative partners (commitment B = — 0.185, p < .001;
predictor, and one with both factors as predictors.
satisfaction B = -0.176, p < .001). We used Cramer's (1972)
*/i<.05. "p<.01.
model comparison procedures to determine whether this two-
factor model predicted evaluations better than did either single-
factor model. The two-factor model was significantly more pow-
tions of alternative partners declined significantly (B = -0.125, erful than either single-factor model (respective Fs = 11.96 and
p < .001), leavers' evaluations of alternative partners actually 51.82), and the commitment variable appears to be the more
increased over time (B = 0.324, p< .001). As expected, the link potent mediator: In predicting evaluations of alternative part-
between time and evaluations of spending time alone was not ners, including satisfaction along with commitment in the re-
significant; evaluations of spending time alone did not decline gression model increased the model's predictive power by only
significantly over time (B = —0.041, p < .187). However, the 0.6% over that provided by commitment alone.4
Leaver X Time interaction was significant (B = 0.272, p <
.024); leavers' evaluations of spending time alone declined sig- 2
The summed commitment variable included one measure that re-
nificantly less than did those of other subjects. fers directly to alternatives to the current relationship: "How attractive
Do these findings represent devaluation of alternatives an alternative would you require before adopting it and ending your
among the highly committed or enhancement of alternatives relationship?" To determine whether these effects occurred simply be-
among less committed persons? As can be seen in Figure 1, both cause items referring to alternatives appeared on both sides of the re-
processes appear to be operating: At Time 1, leavers and stayers gression equation, we calculated a four-item commitment scale, exclud-
rated alternative partners at or just below 5, the scale midpoint. ing the aforementioned measure, and obtained similar findings (for
Over time, stayers' evaluations declined to about 3, whereas evaluations of alternative partners, B = -0.240, p < .001; for evalua-
tions of spending time alone, B - -0.029, p < .293).
leavers' evaluations rose to about 6. As reported earlier, both 3
We performed further analyses, including terms to examine main
slopes differed significantly from 0 (-.125 for stayers vs. .324
effects and interactions involving gender. These analyses revealed that
for leavers). Also, posthoc comparisons revealed that whereas although commitment was significantly negatively related to evalua-
the Time 11-12 ratings of alternative partners provided by stay- tions of alternative partners for both women and men, this effect was
ers were significantly below the scale midpoint, those for leavers stronger among men (evaluations of alternatives for the whole sample,
did not differ significantly from the midpoint (Ms = 2.56 for B = -0.212, p < .001; for the Sex X Evaluations of Alternatives interac-
stayers and 5.68 for leavers). tion, B = -0.127, p < .003). Both slopes are thus negative and differ
significantly from zero (for women, B = -0.212, p < .01; for men, B =
-0.339, p<. 01).
4
The summed measure of evaluations of alternative partners in-
native partners declined over time, women's evaluations declined sig- cluded one item that asked for direct comparisons to the current part-
nificantly more than did those of men (evaluations of alternatives term, ner "In terms of the sorts of rewards and costs discussed above, how
B = -0.305, p < .001; for the Sex X Evaluations of Alternatives interac- difficult would it be to replace your partner?" To determine whether
tion, B = 0.156, p < .026). Both slopes are thus negative and differ sig- evaluations of alternative partners declined over time among the highly
nificantly from zero (for women, B = -0.305, p < .01; for men, B = committed, independent of possible changes over time in feelings re-
-0.149,p<.05). garding the current partner, we repeated all of our analyses, excluding
 RESISTING TEMPTATION 971

6-

5-

4-

3- •*• Stayers-Regress
-n- Leavers-Regress
•*• Stayers-Actual
•*• Leavers-Actual

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

Time Periods
Figure 1. Changes over time in evaluations of alternative partners for stayers and leavers—
actual data and regression (Regress) analysis results: Study 1.

Discussion an active test of this hypothesis, the Study 1 results are merely
suggestive with respect to the validity of this prediction, in that
Thus, over time in a relationship, perceptions of alternatives
we have no evidence that tendencies to devalue alternatives re-
become increasingly less favorable. This change occurs with re-
sult from experienced threat or that the causal ordering is that
spect to evaluations of actual alternative partners, but not for
argued herein. Studies 2 and 3 attempt to address these issues.
evaluations of spending time alone (i.e., the option of noninvolve-
ment). The tendency to evaluate potential alternative partners in
increasingly negative ways is significantly less pronounced among
Study 2
persons for whom the alternative is less threatening: those who even- Study 2 extends Study 1 by exploring devaluation of potential
tually end their relationships. Indeed, among leavers, evaluations of alternative partners among the highly committed under condi-
alternative partners actually become more positive over time. Of tions of high threat: when confronted with the realistic possibil-
course, it is possible that the causal ordering of these factors is re- ity of forming a relationship with a highly attractive alternative
versed. It may be that feelings about noninvolvcment can wax and partner. In this experiment, we used a computer dating service
wane without having any necessary implications for staying in or paradigm to manipulate two independent variables, attractive-
leaving the current relationship, whereas changes in feelings regard- ness of alternative (high, medium, or low) and personal versus
ing actual alternative partners lead more directly to changes in de- impersonal evaluations of the target. Subjects completed a ques-
sire to stick with or end a relationship. tionnaire that obtained information regarding their satisfaction
We also found that although evaluations of alternative part- with and commitment to current relationships and evaluated a
ners are negatively related to both commitment and satisfac- potential date, an "early applicant for the dating service."
tion, it appears that commitment more directly mediates the
tendency to evaluate potential alternative partners negatively.
Method
These results are consistent with our prediction that the highly
committed devalue alternatives as a means of protecting their Subjects. Subjects were 278 undergraduates (117 men and 161
ongoing relationships. However, these findings do not provide women) who participated in partial fulfillment of the requirements for
introductory psychology courses at the University of Kentucky. Partici-
pation was limited to those who were currently involved in heterosexual
dating relationships. We randomly assigned 4-12 same-sex persons re-
this item, and obtained very similar findings (e.g.. examining changes cruited for each session to one of six conditions. The data from 5 non-
overtime,time,B- -0.456,p< .001;leaversB = 0.516,p< .012). White subjects (4 women and 1 man) were eliminated from the analyses
972 DENNIS J. JOHNSON AND CARYL E. RUSBULT

as target dates were White, and we feared that some students might uni- crete aspects of the partner and relationship, 4 items that assessed satis-
laterally reject dates of a different race. faction, 4 items that assessed commitment, and 3 general information
Procedure. Subjects were told that the purpose of the study was to items. The concrete items (e.g., rated intelligence, sense of humor) were
assess attitudes regarding the establishment of a campus-based com- included as a means of involving subjects in the process of thinking
puter dating service. A professor in the psychology department was said about their partners and relationships and are not discussed further.
to have begun collecting applications for such a service while assessing The satisfaction measures were as follows: "In general, to what degree
student interest. Subjects completed questionnaires concerning atti- are you attracted to your partner?" (1 = not at all. 9 = extremely), "In
tudes about a dating service, satisfaction with and commitment to their general, to what extent are you satisfied with your relationship?" (1 =
current relationships, and evaluations of a target date, an early applicant extremely, 9 = not at all; reversed), "All things considered, how strong
for the service. Because the questionnaire that assessed attitudes about are your feelings for your partner?" (1 = love my partner very much,
a computer dating service was included only to increase the credence of 9 - don't love my partner very much; reversed), and "All things consid-
the cover story, it is not described. The order in which subjects com- ered, how does your relationship compare to other people's relation-
pleted the primary tasks—describing their current relationships and ships?" (1 = much better than most, 9 = much worse than most; re-
evaluating the alternative date—was counterbalanced across condi- versed). The commitment measures were "In general, to what extent do
tions. you feel committed to maintaining your relationship?" (1 = extremely,
Each subject was presented with a 3- X 5-in. photo of a fictitious early 9 = not at all; reversed), "In general, for how much longer do you want
applicant for the service, along with a faked application form. The form your relationship to last?" (1 = week or so, 9 = decades), "All things
presented some sketchy personal information about the applicant, in- considered, how likely is it that your relationship will end in the near
cluding self-reported personality traits (11 nine-point bipolar scales) future?" (1 = extremely likely to end, 9 = not at all likely to end), and
and reported interest in a variety of activities (12 nine-point scales). The "All things considered, to what extent do you feel 'attached' to your
applicants were described moderately favorably: as relatively coopera- partner (for better or worse, whether you are satisfied or not)?" (1 =
tive, calm, happy, imaginative, active, good, flexible, relaxed, brave, extremely attached, 9 = not at all attached; reversed). The general infor-
strong, and sincere. Applicants also indicated that they were interested mation items assessed level of involvement (married, engaged, living
together, dating steadily, dating occasionally, dating casually), duration
in the following activities: trying new things, meeting new people, music,
of relationship (fill in), and whether the relationship was exclusive ("do
travel, sports, reading, movies, dancing, parties, dining out, concerts,
you see only each other?"; no or yes).
and picnics. This information was constant across targets. Each subject
The questionnaire that obtained judgments of the target date in-
evaluated only one applicant, a target date of the opposite sex.
cluded 19 concrete evaluations of the target date, three measures of an-
The target date's physical attractiveness was varied to manipulate al-
ticipated satisfaction, and one measure of desire to date the target. The
ternative attractiveness. Thirty-six photographs (18 photographs of
19 specific items—9-point Likert-type scales—were designed to assess
men and 18 of women) from the yearbook of a midwestern university
specific methods of devaluation. To simplify the analyses we divided the
were enlarged to 3- X 5-in. photos. Twenty-four undergraduates (12
items into three categories, roughly following the so-called stage theo-
women and 12 men) rated the attractiveness of each of 9 photos of a
ries of developing relationships (Kerckhoff & Davis, 1962; Levinger &
person of the opposite sex on a 9-point scale (1 = very unattractive, 9 =
Snoek, 1972; Lewis, 1973; Murstein, 1970; Reiss, 1960). The "early-
very attractive). On the basis of these data, we selected photographs of
stage" items obtained ratings of relatively superficial features: evalua-
four highly attractive targets (M = 7.29), four moderately attractive tar-
tions of the alternative's intelligence, sense of humor, and possession of
gets (M = 4.80), and four unattractive targets (M = 1.75; 2 women and
attractive personal qualities, as well as evaluations of how well friends
2 men at each level of attractiveness).' Within each condition, subjects
would like the alternative and how well the alternative would accept
were randomly assigned to one of two opposite-sex target dates.
the subject's friends. The "middle-stage" items obtained ratings of the
Personal versus impersonal evaluation of alternatives was manipu-
relationship that might be formed with the alternative: similarity of atti-
lated by varying the stated purpose of subjects' judgments. All subjects
tudes and interests, how much fun it would be to spend time with the
were told that the researchers needed to obtain ratings of several early
alternative, how reliable and dependable the alternative would be,
applicants for the service. However, subjects in the personal evaluation
whether the alternative would flirt with others, and whether the alterna-
condition were asked to complete the questionnaire indicating how they
tive would be willing to spend time with the subject. The "later-stage"
personally felt about that person. These subjects were told that if they
items concerned issues thought to be related to long-term compatibility:
later decided to participate in the dating service, this information would
would the two have complementary needs, would the alternative live up
be used to match them with potential dates. In the impersonal evalua-
to agreements developed in the relationship, would the alternative be
tion condition, subjects were asked to complete the questionnaire as
sexually faithful, would the alternative treat the subject well, would
they believed the average university student would; that is, they were
there be open communication between the two, would the alternative
asked to report on how the typical undergraduate would feel about the
be easy to confide in, would the alternative be supportive in times of
target. They were told that if they later decided to participate in the
trouble, and would the alternative freely provide emotional support.
dating service, this information would not be used to match them with
The measures of anticipated satisfaction were 9-point Likert-type
potential dates. To reinforce this verbal manipulation, the two otherwise
scales: "In general, to what degree are you attracted to XT' (1 = not at
identical questionnaires were respectively labeled either Personal Atti-
all, 9 = extremely), "In general, how do you think a relationship with
tudes About X or Average Student's Attitudes About X. The manipula- X would compare to your current relationship?" (1 = much better than
tion thus included two features, a manipulation of opportunity to date current relationship, 9 = much worse than current relationship; K-
the target partner in the future and a manipulation of point of view.
We used this manipulation to make the personal evaluation condition
especially potent; however, it is possible that these two conditions !
Furthermore, subjects in Study 2ratedthe applicants they evaluated
differed not only in terms of threat, but also in terms of a more subtle on a 9-point Likert-type scale: "How physically attractive is X?" (1 =
judgmental factor own point of view versus others' point of view. This not at all, 9 = extremely). A three-level (low, moderate, or high attrac-
potential confound should be kept in mind in evaluating the findings. tiveness) analysis of variance (ANOVA) performed on this item revealed
Dependent measures. Subjects' feelings about their current relation- that the three conditions differed significantly in judged attractiveness;
ships were obtained in the Current Dating Relationship Questionnaire. the respective Ms were 2.87, 5.53, and 7.50, Fl,2, 264) = 228.18, p <
This questionnaire included 23 items that assessed feelings about con- .001.
 RESISTING TEMPTATION 973

versed), and "All things considered, to what extent do you think you tual opportunity to date the alternative, but this effect was
would have a satisfying relationship with X?" (1 = not at all satisfying, inconsistently observed (see Commitment X Personal Evalua-
9 = extremely satisfying). The desire to date the target item wasaforced- tions rows of Table 2).
choice scale: "Do you want to go on a date with X?" (1 = yes, definitely,
We calculated commitment slopes for all six experimental
2 = perhaps, 3 = no, definitely not, reversed).
conditions and found that within the personal evaluation of
Reliability and validity of measures. The reliability of the multiple-
highly attractive alternatives condition, four of five commit-
item measures was assessed by calculating alphas for the items associ-
ated with each construct. Sizable coefficients were obtained for the mea- ment slopes were negative and statistically significant (the re-
sures of commitment to current relationship (.82), satisfaction with cur- spective ft were -5.41, -0.64, -3.67, -2.07, and -2.23). It is
rent relationship (.79), and anticipated satisfaction with target date interesting to note that the devaluation of threatening alterna-
(.90), as well as for the early-stage (.63), middle-stage (.58), and later- tives occurred with respect to our measures of early-, middle-,
stage (.69) evaluation of alternative items. Therefore, a single summed and later-stage factors; devaluation processes were not limited
measure of each construct was formed. To assess the validity of our to any one category of measure.
measure of commitment to current relationships, we calculated zero- Not surprisingly, these analyses also revealed that on average,
order correlations between the commitment measure and several more
subjects more favorably evaluated alternatives to the degree that
objective measures of relationship stability. As expected, commitment
the alternative was more attractive: All five coefficients for the
level was significantly correlated with reported level of involvement (r =
alternative attractiveness effect were positive and statistically
.56), reports of how exclusive the relationship was (r = .41), and reports
of relationship duration (r = .29). significant (see Alternative attractiveness rows of Table 2).
There was also weak evidence that on average subjects more
favorably evaluated alternatives when making personal evalua-
Results
tions: All five coefficients for the personal evaluation effect were
Do highly committed persons reject and devalue alternatives positive, although only two were statistically significant (see Per-
under conditions of high threat? We performed a series of re- sonal evaluations rows of Table 2).
gression analyses that included the following terms: commit- Do these findings represent devaluation of alternatives by
ment level, personal versus impersonal evaluation (0 = imper- highly committed persons, or do they represent enhancement
sonal, 1 = personal), and alternative attractiveness (0 = low, 1 = of alternatives by less committed persons? One way to address
moderate, 2 = high). We also included interaction terms repre- this question is to examine the absolute value of subjects' rat-
senting the impact of commitment within the most threatening ings of alternatives, calculated on the basis of unstandardized
conditions: the Commitment X Personal Evaluation interaction coefficients from the regression analyses. First, do less commit-
(Commitment Level X Personal-Impersonal Evaluations) and ted persons bolster alternatives? Within the most threatening
the Commitment X Alternative Attractiveness interaction experimental condition—personal evaluation of highly attrac-
(Commitment Level X Alternative Attractiveness Level).6 We tive alternatives—low commitment subjects' evaluations were
performed separate analyses for five dependent measures. The at or above the midpoint for each scale. However, their scores
results of these analyses are summarized in Table 2, and the were certainly not at the ceiling. Low commitment subjects'
results for the first of these dependent variables—anticipated personal ratings on 9-point scales of highly attractive alterna-
satisfaction with the alternative—are displayed in Figure 2. tives were 6.44 for anticipated satisfaction, 6.36 for early-stage
It is clear that in the absence of threat—that is, before we take variables, 6.24 for middle-stage variables, and 6.51 for later-
into account the Commitment X High Threat interactions— stage variables, and their desire to date the alternative was 2.43
degree of commitment to current partners has little bearing on on a 3-point scale. These ratings hardly seem wildly inflated.
how subjects reacted to potential alternative partners (see Com- Second, do the highly committed devalue alternatives? Calcu-
mitment level rows in Table 2). Only two of these standardized lations based on the regression results reveal that the ratings of
regression coefficients were significant—for evaluations with alternatives provided by the most highly committed persons in
respect to early-stage and later-stage variables—and in both the personal evaluation, highly attractive alternative condition
cases, the impact of commitment on ratings of alternatives was were close to or below the midpoint of each scale and were sim-
positive. ilar to the ratings of low and moderate attractiveness targets.
Is devaluation by those who are highly committed stronger The mean scores of the most committed subjects in the personal
under conditions of greater threat? There are at least two evaluation, highly attractive alternative condition and of the
sources of threat to the stability of a relationship. The first is most committed subjects in the moderate and low alternative
the presence of a very attractive alternative. Consistent with attractiveness conditions were, respectively, 3.49,3.59, and 3.68
predictions, all five standardized regression coefficients for the for anticipated satisfaction; 1.35, 1.40, and 1.45 for desire to
Commitment X Alternative Attractiveness interaction were
negative and statistically significant; the tendency of committed
6
We performed preliminary analyses that included several additional
persons to devalue alternatives was greater to the degree that the
terms: the Commitment X Highest Threat interaction (Commitment
alternative was more attractive (see Commitment X Alternative
X Personal-Impersonal Evaluation X High Alternative Attractiveness),
Attractiveness rows of Table 2). A second threat to the stability
order of evaluations (the counterbalancing factor; 0 = current partner
of a relationship is the presence of a realistic challenger. As pre- evaluated first, 1 = alternative evaluated first), photograph (photograph
dicted, all five coefficients for the Commitment X Personal 0 or 1 within each attractiveness of alternative condition), as well as all
Evaluations interaction were negative, but only three were sta- main effects and interactions involving gender (0 = women, 1 = men).
tistically significant; the tendency of committed persons to de- These analyses revealed few significant effects and no consistent pattern
value alternatives was generally greater when faced with an ac- of results, so these terms were excluded from the main analyses.
974 DENNIS J. JOHNSON AND CARYL E. RUSBULT

Table 2 as we do not have ratings of our targets from impartial, presum-

Evaluations of Alternative Partners as a Function of Personal ably objective judges, it is difficult to determine unequivocally
Versus Impersonal Evaluation, Attractiveness of Alternative, whether the reported interactions occurred because less com-
and Degree of Commitment to Current Relationships: Study 2 mitted persons bolstered available alternatives or because highly
committed persons devalued alternatives.
Multiple regression analyses df P< Impact of satisfaction level. To assess the impact of satisfac-
Anticipated satisfaction with tion with current relationships on evaluations of alternatives,
alternative partner from we added the satisfaction measure—as well as all relevant inter-
Alternative attractiveness 1.983** .446 12.79 5,257 .001 action terms (e.g., Satisfaction X Personal Evaluations, Satisfac-
Personal evaluations 0.230 tion X High Alternative Attractiveness)—to each of the regres-
Commitment level 0.083
sion models described earlier, and compared these results with
Commitment X Personal
Evaluations -0.291* those presented in Table 2 (Cramer, 1972). The addition of sat-
Commitment X Alternative isfaction interaction terms did not significantly improve the
Attractiveness -1.683** prediction of any of the five measures. Adding the simple satis-
Desire to date alternative faction term significantly improved only one of five models—
partner from that predicting anticipated satisfaction (satisfaction ft -
Alternative attractiveness 2.007** .414 10.62 5,257 .001 -0.195, p < .004). Furthermore, if we replicate the analyses
Personal evaluations 0.426**
reported in Table 2, using satisfaction rather than commitment
Commitment level 0.198
Commitment X Personal terms, and then add the commitment terms to each model, the
Evaluations -0.560** addition of commitment terms significantly improves the pre-
Commitment X Alternative diction of all five measures. Thus, as for Study I, it appears that
Attractiveness -1.776** the tendency to devalue alternatives is more directly mediated
Evaluations of alternative by feelings of commitment than by feelings of satisfaction with
partner—Later-stage current relationships.
variables from
Alternative attractiveness 1.915" .335 6.49 5,256 .001
Personal evaluations 0.381** Discussion
Commitment level 0.364**
Commitment X Personal Thus, the impact of commitment level on evaluations of al-
Evaluations -0.456** ternatives was more marked under conditions of greatest threat:
Commitment X Alternative
when confronted with highly attractive alternative partners
Attractiveness -1.695"
rather than when confronted with alternatives of moderate or
Evaluations of alternative low attractiveness and when making personal rather than im-
partner—Middle-stage
personal evaluations. Although there was some evidence that
variables from
Alternative attractiveness 0.820" .171 1.54 5,256 .177 level of satisfaction with current relationships influenced evalu-
Personal evaluations 0.200 ations of alternatives, the tendency to devalue alternative part-
Commitment level 0.146 ners appears to be most directly mediated by variations in com-
Commitment X Personal mitment to current relationships.
Evaluations -0.219
Commitment X Alternative However, because the Study 2 manipulation of personal ver-
Attractiveness -0.690** sus impersonal evaluation included two features—a manipula-
tion of opportunity to date the alternative partner plus a manip-
Evaluations of alternative
ulation of point of view (own versus average student)—and be-
partner—Early-stage
variables from cause the overall impact of this factor was relatively weak, it is
Alternative attractiveness 1.594** .248 3.36 5,256 .006 difficult to interpret our findings regarding the personal versus
Personal evaluations 0.178 impersonal factor. Also, as Studies 1 and 2 examined the effects
Commitment level 0.255*
of commitment as an attribute variable, it is difficult to know
Commitment X Personal
Evaluations -0.215 whether our findings resulted from feelings of threat resulting
Commitment X Alternative from desire to maintain current relationships or from some
Attractiveness -1.436** other personal attribute that is frequently confounded with
commitment. We attempted to address these two problems in
Note. These findings summarize the results of five-factor regression
analyses for each of five dependent variables. Study 3.
*p<.05. **/><.01.
Study 3
date; 6.13, 6.21, and 6.29 for early-stage variables; 5.98, 6.00, Study 3 extends Studies 1 and 2 by actively manipulating sat-
and 6.02 for middle-stage variables; and 6.07,6.14, and 6.22 for isfaction and commitment in a role-playing experiment, thus
later-stage variables. That highly committed persons' evalua- enabling examination of the eflects of variations in commit-
tions of exceptionally attractive alternatives were quite similar ment while holding level of satisfaction with current relation-
to their contemporaries' ratings of low and moderately attrac- ships relatively constant. All subjects were personally con-
tive alternatives suggests some degree of devaluation. However, fronted with an opportunity to become involved with an alter-
 RESISTING TEMPTATION 975

native partner. Subjects role played the part of an essay ship with the current partner. In the high-commitment condition, the
protagonist faced with an opportunity to meet and interact with essays stated that "you enjoy being involved in this exclusive relation-
an attractive alternative partner and indicated degree of attrac- ship . . . find it to be fulfilling in many ways. . . you would like to
tion to and interest in that person. maintain this relationship." In contrast, the low-commitment condition
essays stated that "you don't really want to be involved in an exclusive
relationship . . . believe it would be more fulfilling to date different
Method people. . .you do not want it to be an exclusive relationship." Satisfac-
tion level was manipulated through variations in descriptions of the pro-
Subjects. Subjects were 76 undergraduates (28 men and 48 women)
tagonist's feelings regarding the current partner and relationship. In the
who participated in the experiment in partial fulfillment of the require-
high-satisfaction condition, the essays stated that "Robert is extremely
ments for introductory psychology courses at Illinois State University.
attractive and intelligent and has a very pleasant personality. . .terrific
We randomly assigned the 4-12 persons recruited for each session to
sense of humor. . . enjoys all the same activities as you, so when you
one of four experimental conditions (7 men and 12 women within each
and he go out you always have a good time . . . relationship has been
condition).
very gratifying." In the low-satisfaction condition, the essays stated that
Procedure. Subjects read essays describing fictional situations and
"you don't find Robert particularly attractive . . . [he] does not have
were asked to place themselves in the position of the essay protagonist.
much of a sense of humor . . . does enjoy some of the same activities
Essays read by men and women were identical except for changes in the
as you. . . when you and he go out you usually have a good time. . .
sex of the protagonist, current partner, and alternative partner. Sarah—
relationship has been acceptable."
the protagonist in women's essays—was described as a 21 -year-old col-
lege student who enjoys a reasonably satisfying life both socially and Dependent measures. The questionnaire included 6 seven-point Li-
academically. For the past 3 months Sarah has been dating Robert. kert-type items designed to assess the effectiveness of the manipulations
However, at a social gathering Sarah notices an extremely attractive man of commitment and satisfaction, and 3 items to measure attraction to
looking at her from across the room. From the way he's looking at her, the alternative, as well as 15 five-point Likert scales that measured ten-
Sarah assumes that he's attracted to her. What does Sarah think about dencies to devalue the alternative. The commitment manipulation
this situation? Does she want to approach this stranger? Does she want checks were "To what extent are you committed to maintaining your
him to approach her? The essay ended with the following sentence: relationship with Robert?" (1 = extremely, 7 = not at all; reversed),
"Think about this situation for a few minutes, reading this essay a sec- "How much longer do you want your relationship with Robert to last?"
ond time if necessary, and then complete the questionnaire." (1 = a very short lime, 1 = a very long time), and "How attached are
Commitment level was manipulated through variations in descrip- you to your relationship with Robert?" (1 = extremely, 7 = not at all;
tions of the protagonist's interest in maintaining an exclusive relation- reversed). The satisfaction manipulation checks were "In general, how

6-

5-

Hi Alt; Pers
4- Hi Alt; Impers
Mod Ale Pers
Mod Alt; Impeis
Lo Alt; Pers
Lo Alt; Impers

Lo Commitment Hi Commitment
Figure 2. Anticipated satisfaction with alternative partners as a function of commitment to current relation-
ships, alternative attractiveness, and personal versus impersonal evaluation—regression analysis results:
Study 2. (Hi Alt = high alternative attractiveness, Pers = personal evaluation, Impers = impersonal evalua-
tion, Mod Alt = moderate alternative attractiveness, Lo Alt = low alternative attractiveness.)
976 DENNIS J. JOHNSON AND CARYL E. RUSBULT

does Robert compare to your ideal partner?" (1 =just like ideal, 7 = nificantly influenced all six measures of attraction to the alter-
not at all like ideal: reversed), "In general, how satisfying is your rela- native, Fs(l, 72) = 27.63, 8.76, 23.65, 18.11,24.43, and 20.67,
tionship with Robert?" (1 = extremely, 1 = not at all: reversed), and "In whereas variations in satisfaction significantly influenced only
general, how attracted are you to Robert?" (1 = extremely, 1 = not at all; three of six measures, fs(l, 72) = 7.43, 0.00. 6.63,0.59,0.10,
reversed). The measures of attraction to the alternative were "Generally,
and 9.71, all interactions ns. However, because our manipula-
how attracted are you to Mr. K?" (the stranger; 1 = extremely, 7 = not
tions were not orthogonal, it is possible that the observed com-
at all; reversed), "Would you approach Mr. K or hope he would ap-
proach you at the social gathering?" (0 = yes, definitely, 1 = no, defi- mitment effects are accounted for by the fact that the commit-
nitely not; reversed), and "Would you want to go on a date with Mr. K?" ment conditions also differed in satisfaction, and it is possible
(0 = yes, definitely, 1 = no, definitely not: reversed). that the satisfaction effects are accounted for by the fact that the
The 5-point Likert scales that assessed devaluation of alternatives satisfaction conditions also differed in commitment. Therefore,
asked subjects to judge the likelihood that the alternative possessed each we changed our approach to the analysis of the data.
of 15 qualities. As in Study 2, these variables were divided into three To control for the confounding of satisfaction and commit-
categories based roughly on extant stage theories of developing relation- ment, we analyzed the data by using regression procedures, in-
ships. The early-stage variables asked subjects to assess the likelihood cluding the satisfaction and commitment manipulation checks
that the alternative was intelligent, had a good sense of humor, had a
as covariates. First, we performed simple regression analyses,
pleasant personality, had an outgoing personality, and was easy to get
including the commitment and satisfaction main effects (0 =
along with. The middle-stage variables asked subjects to assess the likeli-
hood that a relationship with the alternative would possess several quali- low, 1 = high) and the Commitment X Satisfaction interaction
ties: would the two have similar attitudes and interests, would the alter- term. As none of the interaction terms were significant, we
native be comfortable to be around, and would the alternative partner dropped them from the model, and simple main effect models
be honest, sincere, reliable, and dependable. The later-stage items con- were tested. All six models were significant (the respective Rs
cerned issues thought to be related to long-term compatibility: would were .570, .539, .327, .454, .504, and .542). The results of these
the alternative partner be a caring person, be sexually faithful, treat the analyses are summarized in Table 3. These findings are the
subject well, provide emotional support, and be unselfish in a relation- same as those for the ANOVA, as the analyses are identical.
ship.
To assess the impact of commitment independent of the ex-
Reliability of measures. The reliability of the multiple-item measures
tent to which it is confounded with satisfaction, we performed
was assessed by calculating alphas for the items associated with each
regression analyses that included two terms, the commitment
set. We obtained sizable coefficients for the commitment (.81) and satis-
faction manipulation checks (.88) and the sets of items designed to ob- main effect (0 or 1) and the satisfaction manipulation check (ac-
tain early-stage (.47), middle-stage (.60), and later-stage (.73) evalua- tual scores, ranging from 1 to 7). All six models were significant
tions of alternative partners. Therefore, a single averaged measure of (the respective Rs were .602, .514, .327, .457, .503, and .547).
each construct was formed. As displayed in Table 3, even when the effects of satisfaction
were taken into consideration, commitment significantly in-
fluenced all six measures of attraction to the alternative, with
Results
high-commitment subjects reporting significantly lower attrac-
Manipulation checks. A two-factor ANOVA was performed tion to the alternative (see the Commitment covarying satisfac-
on the commitment and satisfaction manipulation checks. tion manipulation check column). To assess the impact of satis-
Compared with low-commitment subjects, those in the high- faction independent of the extent to which it is confounded with
commitment condition reported greater attachment to, com- commitment, we performed parallel analyses that included the
mitment to, and desire to continue their relationships (the re- satisfaction main effect and the commitment manipulation
spective Ms were 3.26 and 5.20), F(l, 72) = 86.56, p < .001. check. Five of six models were significant (the respective Rs
Compared with low-satisfaction subjects, those in the high-sat- were .426, .491, .238, .466, .501, and .559). As summarized
isfaction condition reported greater attraction to their partners, in Table 3, when the effects of commitment were taken into
satisfaction with their relationships, and more favorable com- consideration, satisfaction had a significant negative effect on
parisons to their ideal (the respective Ms were 3.47 and 5.91), only one dependent measure, attraction to the alternative (see
F(\, 72) = 206.62, p < .001. Furthermore, the Satisfaction X the Satisfaction covarying commitment manipulation check
Commitment interaction was not significant for the manipula- column).7 Thus, commitment to relationships is clearly more
tion checks of commitment, F( [ , 72) = 0.75, p < .388, or satis- consistently related to tendencies to devalue alternatives than is
faction, f \ 1,72) = 1.26, p < .266. Unfortunately, our manipula- satisfaction with relationships.
tions were not entirely independent: the commitment variable Do these findings represent devaluation of alternatives
significantly influenced the satisfaction manipulation check,
f\ 1, 72) = 5.60, p < .021, and the satisfaction variable signifi-
7
cantly influenced the commitment manipulation check, F(l, We performed additional regression analyses that included (a) the
72) = 4.92, p < .030. These effects are relatively weak in com- commitment and satisfaction main effects (0 or 1) plus the satisfaction
manipulation check, and (b) the satisfaction and commitment main
parison to the effect of each independent variable on its respec-
effects (0 or 1) plus the commitment manipulation check. These analy-
tive manipulation check and are about equal in strength. Never-
ses revealed the same pattern of results: All six of the analyses of com-
theless, because our manipulations were not entirely orthogo-
mitment effects revealed significant effects of commitment, and only
nal, we modified our data analysis strategy.
one of six analyses of satisfaction effects revealed a significant effect of
Do highly committed and highly satisfied persons devalue al- satisfaction. Furthermore, we performed analyses that included main
ternatives? We initially analyzed our data by using analysis of effect and interactions for subject sex, and none of these analyses re-
variance and discovered that variations in commitment sig- vealed significant effects involving the subject sex variable.
 RESISTING TEMPTATION 977

Tables
Evaluations of Alternative Partners as a Function of Satisfaction Level and Commitment Level: Study 3

Two-factor regressions Commitment Satisfaction

covarying satisfaction covarying commitment
Dependent measure Commitment 0 Satisfaction g manipulation check manipulation check

Attraction to alternative -.506** -.262** -.534" -.364**

Desire to approach -.327** .000 -.327** .072
Desire to date -.476** -.252** -.459** -.124
Later-stage variables -.447** -.080 -.438** .059
Middle-stage variables -.503** -.033 -.500** .119
Early-stage variables -.447** -.306** -.419" -.164+

Note. For the two-factor regressions, table values are the standardized coefficients for each independent variable. The fourth column lists the stan-
dardized coefficients in analyses of commitment effects controlling for variations in degree of satisfaction (i.e., including the satisfaction manipulation
check). The last column lists standardized coefficients in analyses of satisfaction effects controlling for variations in degree of commitment (i.e.,
including the commitment manipulation check).
*p<.05. **p<.01.

among the committed, or do they represent enhancement of it is notable that low-commitment subjects' ratings were not
alternatives among the less committed? We addressed this ques- higher (i.e., closer to the ceiling of each scale), and it is impres-
tion by examining the absolute value of subjects' assessments sive that high-commitment subjects' ratings were generally at
of potential alternative partners. (Figure 3 displays means for or below the scale midpoints. However, as we do not have ratings
the first of six measures.) Among low-commitment subjects, all of the alternative partner from impartial judges, it is impossible
six ratings of alternatives were above the midpoint for each scale to determine unequivocally whether the reported effects oc-
(4.07 on a 7-point scale; 0.92 and 0.79 on 0-1 scales; 4.10,3.32, curred because less-committed persons bolstered alternatives or
and 3.39 on 5-point scales). Among high-commitment subjects, because highly committed persons devalued alternatives.
five of six ratings of alternatives were below the midpoint for
each scale (2.56 on a 7-point scale; 0.46 and 0.48 on 0-1 scales; Discussion
2.87,2.75, and 3.65 on 5-point scales). When one considers that Thus, in the Study 3 simulation experiment, persons who
the alternative partner was described as "extremely attractive," were more committed to their current relationships reported

5-

Hi Satisfaction
Lo Satisfaction

Lo Commitment Hi Commitment
Figure 3. Attraction to alternative partners as a function of commitment
and satisfaction levels—actual data: Study 3.
978 DENNIS J. JOHNSON AND CAPYL E. RUSBULT

lower attraction to the alternative partner, lower desire to date ates tendencies to devalue alternatives, but that level of satisfac-
the alternative, lower interest in approaching or being ap- tion may also play some role in this process.
proached by the alternative, and lower evaluations of the alter- Several strengths and limitations of the present work should
native with respect to early-, middle-, and later-stage variables. be noted: The primary strength of Study 1 is that it examined
These effects were weaker for the satisfaction variable; only one changes over time in evaluations of alternative partners; real
of six dependent variables revealed significant effects as a func- changes in ongoing relationships were explored. The primary
tion of variations in satisfaction level. Thus, as in Studies 1 and strength of Study 2 is that it examined reactions to threatening
2, it once again appears that tendencies toward devaluation of alternatives in a very realistic setting, the context of a computer
alternatives are more powerfully linked to feelings of commit- dating service. That persons who were committed to real, ongo-
ment to ongoing relationships than to feelings of satisfaction. ing relationships devalued attractive and threatening alterna-
tives in such a context increases confidence in the obtained
findings. The primary strength of Study 3 is that it came as close
General Discussion as is possible to the ideal of the laboratory experiment and thus
provided good evidence regarding the causal link between com-
The present research was designed to demonstrate that in mitment and tendencies to devalue alternatives. Given that one
comparison to less-committed persons, individuals who are cannot actively manipulate individuals' commitment in ongo-
highly committed to their current partners reject and devalue ing relationships, the simulation experiment may be the closest
alternative partners, particularly under conditions of high feasible approximation of causal evidence.
threat. Consistent with predictions, Study 1 demonstrated that However, for all three studies we could construct plausible
in comparison to persons who ultimately ended their relation- alternative explanations for the obtained findings. First, it could
ships, those whose relationships persisted evidenced an increas- be that in Study 1, alternative partners actually became less fa-
ing tendency to evaluate alternative partners negatively. As ex- vorable over the course of a relationship (e.g., alternatives may
pected, this tendency was most directly mediated by feelings of have taken themselves out of the running). However, this inter-
commitment. However, although the Study 1 findings are con- pretation seems less plausible as we observed the same phenom-
sistent with the assertion that the highly committed devalue al- enon in Studies 2 and 3, where alternatives did not become less
ternatives as a response to experienced threat, this study does attractive over time. Second, it could be that in Studies 1 and
not provide a direct test of this hypothesis. Therefore, Study 2 2—where commitment was explored as an attribute variable—
effected two manipulations of threat to current relationships. the observed effect was produced by another personal attribute
The Study 2 dating service experiment demonstrated that in that is confounded with commitment (e.g., an openness to inti-
comparison to less-committed persons, the highly committed macy, a longstanding habit of attachment, general lack of inter-
reported lower attraction to alternatives. As predicted, this ten- est in dating around). However, given that we observed the same
dency was greatest when the alternative was very attractive and phenomenon in Study 3—where commitment level was actively
subjects were faced with actual opportunities to date that per- manipulated—this interpretation seems somewhat less plausi-
son. However, the Study 2 manipulation of personal versus im- ble. Third, it could be that in Study 2, committed subjects di-
personal evaluations of alternatives varied both opportunity to minished their evaluations of alternative partners as a deliberate
date the alternative and point of view. Also, as in Study 1, in strategy to lessen the likelihood that they would be paired with
Study 2 commitment was explored as an attribute variable. The a partner from the dating service: not meeting is an excellent
Study 3 simulation experiment extended the findings of Studies way to avoid temptation. Alternatively, rather than reacting to
1 and 2 by demonstrating that when level of satisfaction was the experience of high threat, committed persons could have
controlled, evaluations of attractive alternatives were lower merely been reporting that they realistically viewed the alterna-
among persons who were strongly committed to their current tive as an unpromising person to become involved with, espe-
relationships; that is, variations in commitment influenced cially if they had concerns about their own value as a romantic
evaluations of alternative partners independent of degree of sat- partner. It could even be that committed subjects in Study 2
isfaction with the current relationship. Thus, the present re- were showing generosity: By providing low evaluations of alter-
search provides very good support for our hypotheses. natives, they effectively boosted others' chances of being paired
In all three studies, we also examined the effects of satisfac- with the alternative and maximized the odds that the alternative
tion on devaluation of alternatives. All three studies provided would be paired with someone who was actively interested in
some evidence of a link between satisfaction and tendencies to forming a relationship. However, the fact that Study 3 subjects
negatively evaluate alternatives, but in all three studies this link evidenced the same behavior makes this interpretation less con-
was stronger for commitment than for satisfaction. One is vincing: in Study 3 there was no need to persuade a dating ser-
tempted to conclude that commitment is the more powerful vice not to pair oneself with the alternative, and there were no
source of tendencies to devalue threatening alternatives. How- visible (and potentially more interested and needy) competitors
ever, such a conclusion must be tempered by recognition that for the affections of the alternative. Fourth, it could be that in
the more consistent and powerful effects of commitment may Study 3, our role-playing subjects merely behaved in a way that
have resulted from differential reliability in the two variables in is consistent with stereotypes regarding appropriate behavior in
Studies 1 and 2, from differential variability in the two variables committed relationships. However, in light of the consistency of
in Studies 1 and 2, or from differentially powerful manipula- these findings with those of Studies 1 and 2, this interpretation
tions of the two variables in Study 3. Thus, the most prudent seems somewhat less likely. Ultimately, however, although we
conclusion at present is that commitment fairly clearly medi- can counter each of these alternative explanations, it is clear
 RESISTING TEMPTATION 979

that future research will need to explore the merit of these and consistent with a motivational explanation, but the final answer
other alternative accounts of our findings. may require further empirical work.
Two important questions remain to be addressed. First, with Several directions for future research seem promising: First,
respect to all three studies, we must ask whether the observed it would be interesting to determine whether devaluation of
findings resulted from devaluation of alternatives by highly threatening alternatives is a mechanism aimed primarily at
committed persons or from enhancement by the less commit- keeping one's own mind (and heart) straight, or whether it is a
ted. It could be that low-commitment persons have no barriers means of preparing oneself to actively reject an alternative
against interest in alternative partners and therefore come to suitor. If this phenomenon occurs only when an alternative ac-
idealize alternatives. We tried to address this possibility in all tively pursues the individual, then the process would seem to
three studies, examining the absolute levels of ratings made by serve the latter function (i.e., "if you'll stay away from me, I
low- and high-commitment individuals. On average, low-com- won't put you down"). However, to the extent that this phenom-
mitment persons' evaluations of alternatives were above the me- enon occurs in the absence of a pressing need to push away an
dian of our rating scales, but did not seem especially high; these ardent suitor, the process would seem to serve the function of
ratings were certainly not at the ceiling of our scales. Also, on bolstering feelings of certainty and confidence in the Tightness
average, high-commitment persons' ratings of alternatives were of one's commitment decision. Second, it would be interesting
typically at or below the scale midpoints. Indeed, in Study 2, to explore the limits of this effect. One interesting boundary
high-commitment subjects operating under conditions of high case might be that of open relationships—relationships in
threat rated extremely attractive alternatives at about the same which partners agree that "meaningless" sexual encounters
level as did comparable subjects who rated alternatives of only outside of the relationship are not to be regarded as threats to
moderate and low attractiveness. The consistency of our find- their commitment. Do partners in open relationships manage
ings across a variety of measures using diverse methodologies to pursue encounters with alternatives without rendering those
has convinced us that the best explanation of our findings is the encounters unthreatening, or is devaluation especially pro-
one proffered herein; however, in the final analysis, this question nounced because threat to the current relationship is great?
remains to be further explored in future work. Third, it would be interesting to examine the possibility that
A second important question concerns the motivational un- people assure their partners of their commitment by "defusing"
derpinnings of this phenomenon: Do our findings result from alternatives the partner finds to be threatening—ex-spouses,
commited persons' attempts to deal with the conflict they expe- first loves, attractive coworkers, and so on (e.g., "he was hand-
rience when confronted with a real and tempting challenge to some, but he was a terrible lover"; "she's got a magnificent vitae
their ongoing relationships, as a motivational explanation and great legs, but she has no sense of humor").
would imply (e.g., dissonance theory)? Alternatively, do our
findings stem from a simple perceptual phenomenon whereby
alternatives look less favorable merely because they are being
Conclusions
compared with a very gratifying current relationship, as the
concept of comparison level would imply? This question be-
The current work contributes to the understanding of the
comes especially important when one considers that the Study
maintenance of stable relationships by demonstrating the ten-
2 manipulation of personal versus impersonal evaluation varied
dency of committed persons to view alternatives—especially al-
both threat (i.e., opportunity to date the alternative) and stan-
ternatives who are very attractive, tempting, and available—in
dard of comparison (i.e., own versus others' standards). We do
somewhat less enthusiastic and favorable terms than do less-
not believe that this question can be answered unequivocally
committed persons. These findings suggest that one important
on the basis of the present research, although we have tried to
process by which individuals resist temptation and protect their
provide evidence relevant to this issue. First, Study 2 demon-
strated an interaction of commitment level with alternative at- current relationships may be that of burning their bridges: driv-
tractiveness. If devaluation resulted from a simple perceptual ing away threatening alternatives, or at least driving threatening
phenomenon, shouldn't we have observed a simple main effect alternatives from their minds. This process would seem to serve
of commitment level; shouldn't all alternatives—whether so-so the broader function of enhancing individual well-being by re-
or excruciatingly tempting—be shifted downward n units by ducing internal conflict, in that the end product of this form of
committed persons? In contrast, a more motivational account cognitive activity is that the individual is able to avoid difficult
would argue that tendencies to devalue should be differentially and anxiety-provoking choices. Rather than being faced with
aroused as a function of the degree to which a given alternative the necessity of denying oneself the enjoyment of a very attrac-
is threatening, which is what we observed in our research. Sec- tive alternative partner—or enjoying such an alternative and
ond, let us address findings regarding the relative effectiveness facing possible harm to the current relationship or ending the
of commitment and satisfaction in predicting evaluations of al- current relationship and developing a new relationship with the
ternatives. As the perceptual account proposes that devaluation alternative—individuals are able to achieve peace of mind by
results from unfavorable comparison to a very gratifying cur- simply rendering alternatives harmless and putting potential al-
rent relationship, shouldn't this approach predict a stronger ternative partners out of their minds. Thus, these findings con-
effect of satisfaction than of commitment? In contrast, a moti- tribute to the understanding of the dynamics by which individ-
vational account would propose that devaluation results from uals react to and evaluate potential alternative partners and
threats to feelings of commitment, which is what we observed point to the importance of cognitive processes in the mainte-
in our research. Thus, we believe that our findings are more nance of committed close relationships.
980 DENNIS J. JOHNSON AND CARYL E. RUSBULT

References ment. In R. L. Hamblin & J. H. Kunkel (Eds.), Behavioral therapy

in sociology (pp. 299-322). New Brunswick, NJ: Transaction Books.
Brehm, J. W. (1956). Post-decision changes in desirability of alterna- Levinger, G., & Snoek, J. D. (1972). Attraction in relationship: A new
tives. Journal of Abnormal and Social Psychology, 52, 384-389. look at interpersonal attraction. Morristown, NJ: General Learning
Cohen, J., & Cohen, P. (1975). Applied multiple regression/correlation Press.
analysis for the behavioral sciences. New York: Wiley. Lewis, R. A. (1973). A longitudinal test of a developmental framework
Cramer; E. M. (1972). Significance tests and tests of models in multiple for premarital dyadic formation. Journal of Marriage and the Family,
regression. American Statistician, 26,26-30. 35, 16-25.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Lund, M. (1985). The development of investment and commitment
Stanford University Press. scales for predicting continuity of personal relationships. Journal of
Greenwald, A. G., & Ronis, D. L. (1978). Twenty years of cognitive Social and Personal Relationships, 2, 3-23.
dissonance: Case study of the evolution of a theory. Psychological Re- Murstein, B. I. (1970). Stimulus-value-role: A theory of marital choice.
new, 85, 53-57. Journal of Marriage and the Family, 32,465-481.
Johnson, M. P. (1973). Commitment: A conceptual structure and em- Reiss, I. L. (1960). Toward a sociology of the heterosexual love relation-
pirical application. Sociological Quarterly, 14, 395-406. ship. Marriage and Family Living, 22, 139-145.
Kanter, R. M. (1968). Commitment and social organization: A study of Rusbult, C. E. (1980). Commitment and satisfaction in romantic associ-
commitment mechanisms in Utopian communities. American Socio- ations: A test of the investment model. Journal of Experimental So-
logical Review, 33, 499-517. cial Psychology, 16, 172-186.
Kelley, H. H. (1983). Love and commitment. In H. H. Kelley, E. Ber- Rusbult, C. E. (1983). A longitudinal test of the investment model: The
scheid, A. Christensen, J. H. Harvey, T. L. Huston, G. Levinger, E. development (and deterioration) of satisfaction and commitment in
McClintock, L. A. Peplau, & D. P. Peterson (Eds.), Close relation- heterosexual involvements. Journal of Personality and Social Psy-
ships (pp. 265-314). New York: W. H. Freeman. chology, 45, 101-117.
Kenrick, D. T., & Gutierres, S. E. (1980). Contrast effects and judg- Thibaut, J. W.,&Kelley, H. H. (1959). The social psychology of groups.
ments of physical attractiveness: When beauty becomes a social prob-
New York: Wiley.
lem. Journal of Personality and Social Psychology, 38, 131-140.
Kerckhoff, A. C, & Davis, K. E. (1962). Value consensus and need com-
plementarity in mate selection. American Sociological Review, 27, Received December 7,1987
295-303. Revision received July 18, 1989
Leik, R. K., & Leik, S. A. (1977). Transition to interpersonal commit- Accepted July 19, 1989 •

American Psychological Association Today's Date_

Subscription Claims Information

This form is provided to assist members, institutions, and nonmember individuals with any subscription problems. With the appropri-
ate information provided, a resolution can begin. If you use the services of an agent, please do NOT duplicate claims through them and
directly to us. PLEASE PRINT CLEARLY AND IN INK IF POSSIBLE.

PUNT RILL NAUB OR KEY NAMB OP WSITIUTION MEMBER OR OISTOMER NUMBER (MAY IB FOUND ON ANY PACT HSUB LABEL)

DATS YOUR ORDER WAS MAILED (ORPHONBD)

P. O. NUMBER

STATE/COUNTRY ZIP PHEPAin CHUCK QuROa. CHECK/GAUD CLEARED DAT!

(If possible, send a copy, front and back, of your cancelled check to help us in
YOUR NAME AND PHONE NUMBER our research of your claim.)
Issues; MBSMO __J3uiAaED

TilLB(S) VOU/YR. ISSUES® NO./MONTH

Thank you. Once a claim is received and resolved, delivery of replacement issues routinely takes 4-6 weeks.
_ •.• « on • M •• « nCTo BE FILLED OUT BY APA STAFF).
DATE RECHVED_ DATBOPAcriON_
ACTION TAKBN_ INV. No. & DATB_
STAFF NAMB__ LABEL #, DATB__

SEND THIS FORM TO: APA Subscription Claims, 1400 N. Uhle Street, Arlington, VA 22201
PLEASE DO NOT REMOVE. A PHOTOCOPY MAY BE USED.