Journal of Personality and Social Psychology1983, Vol. 45, No. 1, 101-117

Copyright 1983 by theAmerican Psychological Association, Inc.

A Longitudinal Test of the Investment Model: The Development(and Deterioration) of Satisfaction and Commitment

in Heterosexual Involvements

Caryl E. RusbultUniversity of Kentucky

A longitudinal study of heterosexual dating relationships tested investment modelpredictions regarding the process by which satisfaction and commitment develop(or deteriorate) over time. Increases over time in rewards led to correspondingincreases in satisfaction, whereas variations in costs did not significantly affectsatisfaction. Commitment increased because of increases in satisfaction, declinesin the quality of available alternatives, and increases in investment size. Greaterrewards also promoted increases in commitment to maintain relationships,whereas changes in costs generally had no impact on commitment. For stayers,rewards increased, costs rose slightly, satisfaction grew, alternative quality de-clined, investment size increased, and commitment grew, whereas for leavers thereverse occurred. Individuals whose partners ended their relationships evidencedentrapment: They showed relatively low increases in satisfaction, but their alter-natives declined in quality and they continued to invest heavily in their relation-ships. Suggestive evidence points to the importance of changes over time in com-mitment in mediating stay/leave behaviors. The generalizability of these resultsfor men and women and stayers and leavers at all stages of involvement is dis-cussed.

What causes partners in romantic involve-ments to be satisfied with their relationships?What causes individuals to be committed tomaintaining their involvements? Why dosome relationships persist over time whereasothers end? Most social psychological re-search on such issues has sought to answerthe first of these three questions, that is, toidentify the determinants of attraction andsatisfaction (e.g., Ajzen, 1974; Berscheid,Brothen, & Graziano, 1976; Drachman,DeCarufel, & Insko, 1978; Insko & Wilson,1977; Tyler & Sears, 1977; Walster, Walster,Piliavin, & Schmidt, 1973). Others have di-rected their attention to the second of these

This research was supported by an Institutional Grantfrom Franklin and Marshall College and by a SummerFaculty Research Fellowship from the University of Ken-tucky Research Foundation.

The author is grateful to David Lowery for some ex-tremely useful suggestions regarding the analyses re-ported herein and to Belle Zembrodt for help in codingand checking the data. Special thanks go to the partic-ipants in this study for their unflagging cooperation.

. Requests for reprints should be sent to Caryl Rusbult,Department of Psychology, University of Kentucky, Lex-ington, Kentucky 40506.

questions; they have examined variables pre-dictive of "distress" and/or commitment inongoing involvements (e.g., Billings, 1979;Birchler, Weiss, & Vincent, 1975; Fineberg& Lowman, 1975; Gottman et al., 1976).Finally, some researchers have addressed thethird question, exploring the causes of rela-tionship dissolution (e.g., Bentler & New-comb, 1978; Brown & Manela, 1978; Hill,Rubin, & Peplau, 1976; Kerckhoff, 1976;Levinger & Moles, 1979; Todres, 1978). Thisliterature contributes to our understandingof romantic involvements by identifying avariety of factors that appear to be importantin affecting relationship satisfaction, com-mitment, and dissolution.

Unfortunately, there have been few at-tempts to integrate these diverse findings bydeveloping more abstract, general theories ofromantic involvement. A few authors haveproffered such general models (e.g., Altaian& Taylor, 1973; Hatfield, Utne, & Traupman,1979; Levinger, 1974; Rusbult, 1980a), butmost empirical assessments of these modelshave been "static" in nature. That is, mostresearch designed to evaluate the predictivepower of general theories of romantic in-

101

102 CARYL E. RUSBULT

volvement has done so at a single point inthe development of individuals' relation-ships; very little research has utilized longi-tudinal methodologies. In the absence of lon-gitudinal research, it is difficult to answeradequately a variety of important questionsregarding the process by which relationshipsdevelop and deteriorate over time.

The present research was designed as a lon-gitudinal test of the investment model (Rus-bult, 1980a). The investment model is a sim-ple extension of concepts developed in theexchange tradition within social psychology(Blau, 1964; Romans, 1961;LaGaipa, 1977),particularly interdependence theory (Kelley& Thibaut, 1978; Thibaut & Kelley, 1959).As in interdependence theory, the investmentmodel distinguishes between two importantcharacteristics of relationships: satisfaction—positivity of affect or attraction to one's re-lationship—and commitment—the tendencyto maintain a relationship and to feel psy-chologically "attached" to it. This definitionof satisfaction corresponds to that utilized byother authors (i.e., generally having positivefeelings about one's partner/relationship),whereas the definition of commitment in-cludes two categories of definition advancedby other authors: behavioral intent and psy-chological attachment (cf. Johnson, 1973;Rosenblatt, 1977). From the point of view ofthe investment model, these two types ofcommitment should covary. That is, whetheror not they are satisfied, persons who reportintent to maintain their involvements shouldalso report feelings of psychological attach-ment. The model asserts that variations inlevel of commitment should mediate stay/leave decisions.

Following interdependence theory, the in-vestment model argues that individuals shouldbe more satisfied with their relationships(SATX) to the extent that they continue toprovide high rewards (REWX) and low costs(CSTJ and exceed their generalized expec-tations, or comparison level (CLx). Thus, ifindividuals share many common interestswith their romantic partner (i.e., derive nu-merous rewards) with whom they seldomargue (i.e., incur few costs), and expect littlefrom their romantic involvements more gen-erally (i.e., have a low comparison level), thenthey should be relatively satisfied with their

involvement. These relationships are ex-pressed as follows:

SATX = (REWX - CSTX) -

Greater satisfaction with an involvementshould increase commitment to maintainthat relationship (COMX). However, commit-ment should also be influenced by two ad-ditional variables: alternative quality and in-vestment size. As in interdependence theory,the investment model proposes that personsbecome more committed when they perceivethat they have only poor alternatives to theircurrent associations. Perceived quality of al-ternatives (ALTX) is established in much thesame manner as is satisfaction with currentinvolvements (i.e., it is influenced by the an-ticipated rewards and costs of the alterna-tive), except that in this case the object ofevaluation is an alternative relationship, sol-itude, "dating around," or spending time withfriends and relatives (i.e., whatever is the bestavailable alternative). For example, if indi-viduals are relatively dissatisfied with theirrelationship (i.e., experience low satisfaction)and really enjoy spending time on their own(i.e., have a good alternative), they should beless committed to maintaining their relation-ship.

According to the investment model, a thirdmeans of becoming committed to a relation-ship is by investing numerous resources inthat involvement. There are two general cat-egories of investment (INVX): intrinsic andextrinsic. Intrinsic investments are those re-sources that are put directly into the rela-tionship, such as time, emotional effort, orself-disclosures. Extrinsic investment occurswhen initially extraneous resources becomeinextricably connected to the relationship(e.g., mutual friends, shared memories ormaterial possessions, activities/persons/ob-jects/events uniquely associated with the re-lationship). Invested resources may also proveto be rewarding or costly, for example, sharedmemories or mutual friends could also serveas rewards, whereas emotional effort or mon-etary investments could be costly. What dis-tinguishes between rewards/costs and invest-ments is that once invested, intrinsic and ex-trinsic investments cannot readily be removedfrom the relationship; normal rewards and

INVESTMENT MODEL 103

costs are not as strongly tied to a particularinvolvement and do not as dramatically de-cline in value (or become lost) with the dis-solution of a relationship. Investments in-crease commitment and help to "lock the in-dividual into his or her relationship" byincreasing the costs of ending it—to a greateror lesser degree, to abandon a relationship isto sacrifice invested resources. The presentuse of the investment concept, thus, is similarto constructs advanced by other authors:Becker's (I960) notion of "side bets" (i.e.,extrinsic investments), Levinger's (1979) dis-cussion of "barrier forces," or issues relatedto entrapment and investments as presentedby Rubin (Rubin & Brockner, 1975), Blau(1964), and Staw (Staw, 1976; Staw & Fox,1977).

Thus, according to the investment model,an individual's commitment to maintain arelationship should increase to the extent thathe or she is satisfied with that involvement,has no acceptable alternative, and has in-vested in it heavily. This proposition may beexpressed as follows:

COMX = SATX- ALTX + INVX. (2)

If we ignore some complications involvingcomparison level, these relations may be al-ternatively expressed as

COMX = (REWX - CSTX)

-ALTX + INVX. (3)

Stay/leave behaviors (ST/LVX) are said to bedirectly mediated by the individual's psycho-logical/cognitive commitment to maintainhis or her relationship:

ST/LVX = COMX. (4)

Because satisfaction and commitment neednot necessarily be strongly correlated—strongcommitment may be produced by poor al-ternatives or large investments—it is possibleto be dissatisfied with a relationship yet re-main committed to it and stay involved init. Alternatively, an individual might leave arelatively satisfying involvement because ofthe availability of an attractive alternativecoupled with low investments in the currentrelationship.

The predictive power of the investmentmodel has heretofore been explored in ro-

mantic involvements, friendships, and busi-ness associations. I (Rusbult, 1980a) assessedthe predictive ability of the investment modelin two studies of romantic associations—arole-playing experiment and a cross-sectionalsurvey—and (Rusbult, 1980b) employed across-sectional survey to examine the predic-tive ability of investment model variables infriendships. Finally, several studies have ap-plied the investment model to the study ofjob satisfaction, job commitment, and turn-over: Farrell and Rusbult (1981) reported anexperimental simulation of a work settingand a cross-sectional survey of industrialworkers, whereas Rusbult and Farrell (inpress) presented a longitudinal study ofprofessional workers. This research providesconsistently good support for the model.

However, although the investment modelwas initially proposed as a means of account-ing for the development and deterioration ofsatisfaction and commitment, and as a meansof accounting for actual stay/leave behaviors,such phenomena have not yet been directlyexamined in romantic relationships. As notedabove, previous investment model researchin the domain of romantic involvement hasbeen static rather than longitudinal. The pres-ent research attempts to remedy this defi-ciency by exploring both the processes by

^ which satisfaction and commitment "ebband flow" over time and the dynamics ac-counting for the termination of involve-ments.

More specifically, the present research wasdesigned to address three basic issues. Thefirst issue at hand was to assess the overallpower of investment model variables in pre-dicting satisfaction and commitment: Acrossthe entire sample, does the model influencesatisfaction and commitment as specified inEquations 1, 2, and 3 above? Does the model"behave" similarly for men and women, forstayers and leavers, and at all stages of in-volvement? A second goal was to explore themanner in which investment model variableschange over time. For the sample as a whole,Are there significant trends over time in re-ported rewards, costs, alternatives, invest-ments, satisfaction, and commitment? Aresuch changes similar for men and women?Finally, and perhaps most importantly, thestudy sought to establish the relationship be-

104 CARYL E. RUSBULT

tween investment model variables and stay/leave behaviors. Two related goals were toevaluate the relationship between "intent"commitment and "attachment" commit-ment and to assess the role of commitmentin mediating stay/leave behaviors (Equa-tion 4).

Thirty-four individuals participated in the7-month longitudinal study of romantic in-volvements. Every 17 days subjects com-pleted questionnaires designed to measurerewards, costs, alternatives, investments, sat-isfaction, and commitment.1 Subjects' par-ticipation concluded when their relationshipsended or when the study itself ended. Duringthe course of the 7-month study, 29% of thesubjects' relationships terminated (10 out of34 relationships).

Method

Subjects and Procedure

Subjects were 17 male and 17 female undergraduatesfrom Franklin and Marshall College. Individuals vol-unteered to participate in the study in response to a flyerplaced in the campus mailbox of each student at thecollege. The flyer described the purpose of the study,requirements for eligibility (heterosexual, unmarrieddating couples), nature of requests that would be madeof participants, and instructions for volunteering. Onehundred and nineteen individuals volunteered to par-ticipate. Subjects were selected based on sex (half men,half women) and initial duration of the relationship (fivemen and five women in the 0-2-week category; four ofeach in the 2-4-week, 4-6-week, and 6-8-week catego-ries). Only one partner in a given relationship was al-lowed to participate. Beyond these requirements, partic-ipants were selected on a first come, first served basis.The mean age of subjects was 20.14 years for men and19.44 years for women. The mean duration of relation-ships at the start of the study was 4.24 weeks for menand 4.06 weeks for women. Participants were paid $2.50for each questionnaire they completed, each of whichrequired approximately 25 minutes to answer.

The study began near the start of the academic year,and subjects' participation concluded when their rela-tionships ended or the study itself ended (at the conclu-sion of the academic year). One man dropped out of thestudy at Time 2, and two men and one woman stoppedresponding at Time 8 (spring break). Of the remaining30 subjects, 10 persons' relationships ended (four men,six women), and 20 subjects responded throughout thestudy (their relationships extended beyond Time 12; 10men, 10 women). Participants returned questionnairesthrough campus mail within 4 days of their receipt. Ifa questionnaire was not returned within 7 days of thetime it was mailed, a reminder was mailed to the par-ticipant in an attempt to encourage him or her to com-plete it promptly. This procedure was altered only twice

during the course of the study—during winter break(1978), when questionnaires with return envelopes weremailed to participants' off-campus addresses, and during

xthe Three Mile Island incident (1979), when the collegewas closed for 1 week (questionnaires were delayed 4days). Participants exhibited remarkable responsibilitythroughout the study; most questionnaires were quicklycompleted, and relatively few reminders were required.At the end of the study, all participants were mailed fulldescriptions of the purpose of the study, were given a setof simple descriptive data presenting preliminary find-ings, and were thanked for their unflagging cooperation.

QuestionnairesSubjects completed two types of questionnaires: (a)

an Initial Questionnaire, which requested that partici-pants describe their perceptions of their relationships attheir very start, and (b) the 12 identical RelationshipQuestionnaires completed every 17 days, which com-prised the bulk of the study's data and were designed tomeasure the major parameters of the investment model.Unless otherwise indicated, measures were 9-point scaleswith bipolar adjectives/phrases as end anchors.

Initial Questionnaire. The Initial Questionnaire and12 Relationship Questionnaires were designed to mea-sure each variable of the investment model—rewards,costs, alternatives, investments, satisfaction, and com-mitment—and were modeled after the questionnaireused in Rusbult (1980a). Each concept was briefly de-fined, subjects answered a series of items representingconcrete operationalizations of each model variable, andthen subjects completed general measures tapping eachvariable. The concrete measures prepared subjects toanswer the more general questions and, in a sense,"taught" them the meaning of the general item. Themeans and standard deviations of each general measureare presented in Table 1.

The following issues were addressed in the concretemeasures designed to assess initial rewards and costs:partner's personality, partner's attitudinal similarity,partner's intelligence, partner's physical appearance,partner's physical proximity, partner's similarity of val-ues, partner's need complementarity, partner's sense ofhumor, partner's shared interests, partner's similarity ofhabits or pastimes, partner's sexual relationship, sub-ject's loss of personal freedom, relationship's monetarycosts, relationship's time costs, partner's embarassinghabits, partner's unattractive personal qualities, partner'sunattractive attitudes about relationships, and partner'sreliability. One general item measured initial relation-ship rewards ("Initially, how rewarding was the relation-ship?"; 1 = not at all, 9 = extremely), and one generalitem measured costs ("Initially, how costly was the re-lationship?"; 1 = not at all, 9 = extremely). Two general

' I did not attempt to measure comparison level inthis study because of the inextricable connection be-tween individuals' reports of rewards and costs and theirgeneralized expectations. That is, I expected that it wouldbe very difficult for participants to separate that whichexists objectively from that which they expect more gen-erally.

INVESTMENT MODEL 105

questions assessed alternatives: "In terms of the sorts offactors discussed above, how appealing are the peopleyou could have dated other than your partner?" (1 =not at all, 9 = extremely) and "To what extent can yoube happy when you're not dating anyone?" (1 = not atall happy, 9 = very happy). The concrete items measuringinvestments were mutual friends, duration of acquain-tance, hours per week spent together, shared "memories"or events, monetary investments, shared possessions,activities uniquely associated with relationship, emo-tional investment, and self-disclosures. The two generalmeasures of initial investment size were "Initially, howmuch had you invested in the relationship?" (1 = verylittle, 9 = very much) and "At that point, what wouldyou have lost by ending the relationship?" (1 = lost agreat deal, 9 = lost nothing). The general measures ofinitial satisfaction were "Initially, to what extent wereyou attracted to your partner?" (1 = not at all, 9 =extremely) and "Initially, to what degree were you sat-isfied with your relationship?" (1 = extremely, 9 = notat all). Three general items measured initial commit-ment: "Initially, for how long did you want your rela-tionship to last?" (1 = week or so, 9 = lifetime). "Towhat extent were you committed to maintaining the re-lationship?" (1 = extremely, 9 = not at all), and "Towhat extent were you 'attached' to him/her" (1 = notat all, 9 = extremely).

Relationship Questionnaires. The 12 RelationshipQuestionnaires, too, obtained concrete and general mea-sures of all investment model variables. The concreteitems are described below, and the general measures arelisted in Table 1 along with the mean and standard de-viation of each measure at each time. The following fac-tors were assessed in the concrete measures of rewardsand costs: partner's need complementarity, partner's at-titudinal similarity, partner's similarity of values, part-ner's physical appearance, partner's personality, partner'sintelligence, partner's sense of humor, partner's sharedinterests, partner's physical proximity, partner's sexualrelationship, partner's similarity of interests or pastimes,partner's ease of communication, partner's reliability,subject's loss of personal freedom, relationship's mon-etary costs, relationship's time costs, partner's embar-assing habits, partner's unattractive personal qualities,partner's unattractive attitudes about relationships, part-ner's failure to live up to agreements developed in re-lationship, conflict in relationship, partner's sexual faith-fulness, reciprocation in relationship, dependency in re-lationship, partner's feelings of guilt, partner's emotionalstability, and emotional "ups and downs" in relationship.The following were tapped in the concrete measures ofalternatives: appeal of alternative relationships, difficultyof replacing partner, importance of involvement, andsubject's happiness when not involved. The concretemeasures of investments were number of hours per weekspent together (fill in), exclusivity of involvement, mu-tual friends, shared "memories" or events, monetary in-vestment, shared possessions, activities uniquely asso-ciated with relationship, emotional investment, and self-disclosures. Satisfaction and commitment were mea-sured only with general items (see Table 1). Subjects werealso asked to write brief descriptions of anything unusualthat had happened in their relationships since complet-ing the previous questionnaire. If their relationshipsended, they were asked to describe the cause(s) of the

termination and the circumstances surrounding theirbreakup (e.g., Who ended it? Why do you think itended?).

Results

Reliability of Measures

Reliability estimates (coefficient alpha)were computed at each time for the set ofgeneral measures designed to estimate eachmodel variable. (For the Initial Question-naire, these analyses could not be performedfor the measures of rewards and costs becauseeach was assessed by a single general mea-sure.) The results of the reliability analysesare presented in Table 2.. Generally, thesecoefficients exceeded recommended levels(Nunnally, 1967; the only exceptions wereALT and INV at Time 0 and GST at Time1), so a single measure of each investmentmodel variable, at each time, was formed byaveraging the individual general measures ofeach variable. These averaged measures wereused in the remaining analyses.2

Predicting Satisfaction and Commitment

The data for the following analyses wereaveraged measures of each investment modelvariable for each of 34 subjects at each of 13times (Time 0 through Time 12, or until thesubject's relationship ended). Usual multipleregression analyses could not be applied tothese data, because the 13 sets of measuresfor a given subject cannot be assumed to beindependent observations. To account for theportion of the variance controlled by thewithin-subjects factor, a subjects variable(SUBJ) was included as a categorical factorin all regression analyses using techniquesoutlined in Cohen and Cohen (1975).

Overall sample. According to Equation1 of the investment model, rewards shouldbe positively related to satisfaction and costsshould be negatively related to satisfaction.Therefore, the measures of rewards and costswere regressed onto corresponding satisfac-

(text continued on page 108)

2 Measures were scored in reverse where necessary sothat higher numbers always indicated greater rewards,costs, satisfaction, alternatives, investments, and com-mitment.

Table 1Means and Standard Deviations of All General Measures at Times 0 Through 12

General measure

RewardsHow rewarding is this relationship? (1 = not at all,

9 = extremely)

In terms of rewards, how does the relationshipcompare to your ideal? (1 -= close to ideal,9 = far from ideal)

M

CostsHow costly is this relationship? (1 = not at all,

9 = extremely)

In terms of costs, how does the relationshipcompare to your ideal? ( 1 = close to ideal,9 = far from ideal)

M

AlternativesIn general, how appealing are your alternatives

(dating another person or other persons, orbeing without a romantic involvement)? (1 =not at all appealing, 9 = extremely appealing)

All things considered, how do your alternativescompare to your current relationship? (1 =this is much better, 9 = alternatives are muchbetter)

M

Investment SizeAll things considered, are there objects/persons/

activities associated with the relationship that

0

7.38(1.28)

7.38(1.28)

2.82(1.73)

2.82(1-73)

4.97(2.47)

6.12(2.47)

5.59(1.61)

3.94(2.45)

1

8.03(1.24)

7.29(1.68)

7.66(1.35)

2.50(1.50)

3.65(2.31)

3.07(1.62)

3.65(2.01)

2.29(1-64)

2.97(1.66)

5.44(2.48)

2

8.09(1.07)

7.52(1.30)

7.80(1.11)

X73(1.83)

3.70(1.85)

3.21(1.67)

'.

3.33(2.03)

2.15(1.30)

2.74(1.50)

6.24(2.19)

3

8.00(1.00)

7.36(1.47)

7.68(1.17)

3.16(1.99)

3.52(2.05)

3.34(1.86)

3.26(1.83)

2.10(1.30)

2.68(1.44)

6.43(2.03)

4

7.90(1.14)

7.36(1.87)

7.63(1.44)

3.45(2.00)

3.61(2.11)

3.53(1.94)

3.03(2.09)

2.45(1.96)

2.74(1.90)

6.45(2.20)

5

7.87(1.33)7.17

(1.80)

7.52(1.48)

3.33(2.19)3.67

(2.12)

3.50(1.99)

3.18(2.23)

•2.45(2.03)

2.84(2.07)

6.67(2.40)

Time

6

8.00(.86)

7.76(1-02)

7.88(-90)

3.48(1.92)

3.86(2.55)

3.67(2.17)

3.35(2.37)

2.24(1.43)

2.79(1.80)

6.59(2-11)

7

7.86(1.11)7.52

(1.50)

7.70(1.23)

3.36(2.06)

3.75(2.37)

3.55(2.14)

3.50(2.27)

2.18(1.81)

2.84(1.85)

6.82(1.77)

8

7.75(1.36)

7.42(1.56)

7.58(1.39)

3.30(2.16)3.71

(2.48)

3.44(2.23)

3.17(2.28)

2.92(1.92)

2.73(2.01)

7.21(1.74)

9

7.71(1-52)

7.42(1-59)

7.56(1.50)

3.63(2.26)

3.88(2.23)

3.75(2.13)

3.29(2.39)

2.46(2.09)

2.88(2.10)

7.25(1-85)

10

7.91(1.28)7.44

(1.62)

7.67(1.32)

3.74(2-24)

4.00(2.52)

3.87(2.28)

3.26(2.56)

2.65(1.95)

2.96(2.07)

7.44(2.02)

11

8.05(-95)

7.32(1-52)

7.68(1-12)

3.73(2.39)3.77

(2.60)

3.75(2.41)

3.32(2.66)

2.64(2.28)

2.98(2.41)

7.59(1.71)

12

8.11(.90)

7.67(1.41)

7.89(1.09)

3.00(1-77)

3.28(2.42)

3.09(1.99)

2.83(2.62)

2.39(2.40)

2.61(2.43)

7.94(1.21)

f-\9%rM

JO

ondaGrH

yon would lose (or value less) if therelationship were to end? (1 = none, 9 =many)

Table 1 (continued)

General measure

In general, what is the size of your investment inyour relationship? (1=1 have invested agreat deal, 9 = 1 have invested nothing)

M

SatisfactionHow much do you like your partner? (1 = very

much, 9 = not at all)

To what extent are you attracted to your partner?(1 = not at all, 9 = extremely)

To what degree are you satisfied with yourrelationship? ( 1 = extremely, 9 = not at all)

M

CommitmentHow likely is it that you will end your relationship

in the near future? (1 = not at all likely,9 = extremely likely)

For what length of time would you like yourrelationship to last? (1 = week or so, 9 =lifetime)

How attractive an alternative would you requirebefore adopting it and ending yourrelationship? (1 = extremely attractivealternative, 9 = moderately attractivealternative)

To what extent are you "attached" to yourpartner? (1 = not at all, 9 = extremely)

To what extent are you committed to yourrelationship? (1 = extremely, 9 = not at all)

M

N

0

4.53(2.86)

4.24(2.12)

7.27(1.54)

7.62(1-30)

7.44(1.22)

4.76(1.99)

4.59(2.36)

3.59(2.44)

4.25(1-84)

34

1

6.32(1.93)

5.88(1.99)

8.18(l.li)8.21

(1.10)7.82

(1.22)

8.07(1.19)

7.50(1.99)

5.81(2.40)

6.82(2.29)

8.03(1-45)

7.55(1.80)

7.17(1.76)

34

2

7.22(1.60)

6.80(1.70)

8.33(1.05)

8.21(-93)7.70

(1.29)

8.08(1.18)

7.66(2.03)

5.88(2.45)

7.19(2.18)

8.27(1.57)

7.59(1.97)

7.26(1.78)

33

3

7.23(1.61)

6.83(1.69)

8.32(1.14)

8.32(.98)

7.68(1.45)

8.11(1-29)

7.81(1.78)

5.93(2.35)

7.23(1.86)

8.07(1.46)7.94

(1.55)

7.38(1.49)

31

4

7.45(1-52)

6.95(1.64)

8.36(1.25)

8.36(LOS)

7.65(1.60)

8.12(1.45)

7.65(2.06)

5.63(2.34)

7.29(2.13)

8.26(1.53)7.77

(1-84)

7.30(1.62)

31

5

7.60(1.75)

7.13(1.89)

8.20(1-32)

8.00(1.74)

7.37(2.21)

7.86(2.04)

7.20(2.28)

5.55(2.49)

7.50(1.74)

8.13(1.43)

7.90(1.69)

7.26(1.64)

30

Time

6

7.79(1.52)

7.19(1.62)

8.52(.69)

8.41(.83)7.66

(1-57)

8.20(1.10)

7.62(2.15)

6.00(2.46)

7.62(1.45)

8.17(1.42)

8.03(1.38)

7.48(1.32)

29

7

7.64(1.47)

7.23(1.48)

8.32(-98)

8.32(.95)7.54

(1.80)

8.06(1.63)

7.39(2.15)

6.11(2.27)

7.50(1.73)

7.89(1.91)

7.93(1-49)

7.36(1.55)

28

8

7.71(1.83)

7.46(1.72)

8.21(1.32)

8.29(1.23)7.67

(1.71)

8.06(1.65)

7.67(2.01)

5.96(2.51)

7.13(2.01)

8.13(1.33)

7.75(1.87)

- 7.33(1.71)

23

9

7.54(2.04)

7.40(1.89)

8.21(1.25)

8.17(1.20)

7.58(1.56)

7.99(1.59)

7.29(2.35)

6.00(2.67)

7.25(2.19)

8.04(1.68)7.67

(2.04)

7.25(1.97)

23

10

8.13(1.55)

7.78(1.65)

8.30(1.11)

8.13(1.36)

7.26(1.94)

7.90(1.71)

7.39(2.41)

6.35(2.93)

7.35(2.06)

8.30(1-43)7.83.

(1.95)

7.44(1.96)

22

11

8.20(1.15)

7.90(1.44)

8.18(1.44)

8.05(1.70)7.04

(2.34)

7.76(2.09)

6.96(2.80)

6.00(2.81)

7.05(2.36)

8.18(1.79)7.73

(2-07)

7.18(2.11)

21

12

8.33(-84)

8.14(.94)

8.50(-71)

8.33(.91)7.67

(1.82)

8.17(1.32)

7.56(2.15)

6.94(2-59)

7.83(1-38)

8.33(1.91)8.28

(1.23)

7.98(1-21)

20

Note. Measures were scored in reverse where necessary so that higher numbers indicate greater rewards, greater costs, better alternatives, greater investment size, highersatisfaction, and higher commitment Standard deviations are in parentheses. The general measures listed above are those utilized at Times 1 through 12. The generalmeasures at Time 0, which differ sightly from these, are presented in the Method section. '

108 CARYL E. RUSBULT

Table 2Reliability of Investment Model Measures

Time

0123456789

101112

REW_

.80

.86

.85

.85

.86

.87

.84

.89

.93

.78

.71

.82

CST_

.55

.79

.83

.88

.83

.92

.92

.93

.88

.90

.93

.85

SAT

.64

.81

.86

.83

.87

.92

.88

.79

.93

.96

.92

.90

.83

ALT

.32

.77

.71

.79

.86,94.82.77.90.86

. .79.95.93

INV

.41

.75

.70

.82

.67

.76

.71

.79

.92

.94

.81

.91

.76

COM

.75

.93

.84

.88

.91

.88

.80

.88

.89

.95

.93

.93

.83

Note. Table values are alphas for each analysis. Time 0 =Initial Questionnaire, Times 1-12 = Relationship Ques-tionnaires. REW = rewards, CST = costs, SAT = sat-isfaction, ALT = alternatives, INV = investments, andCOM = commitment.

tion measures (at Time 0 through Time 12).The results of this analysis are displayed inTable 3. The multiple correlation of rewardsand'costs with satisfaction was significant(R2 = .88), but although the regression coef-ficient for the rewards factor was significant(/8 = .890), the coefficient for the cost factorwas not (|8 = .045); rewards contributed sig-

Table 3Multiple Regression Tests of Investment ModelPredictions: Overall Sample

Measure

SAT fromREWCST(SUBJ)

COM fromSATALTINV(SUBJ)

COM fromREWCSTALTINV(SUBJ)

0

.890*

.045

—

.845*-.500*

.840*

—.901*.078

-.595*.720*

—

R2 F

.88 69.02

.89 71.18

.90 76.12

nificantly to the prediction of satisfaction,but costs did not. Equation 2, for commit-ment, was evaluated by examining the mul-tiple regression between the commitmentmeasure and measures of satisfaction, alter-natives, and investments (see Table 3). Aspredicted, satisfaction and investment sizewere positively related to commitment,whereas alternative quality and commitmentwere negatively related. Finally, Equation 3was tested by regressing rewards, costs, al-ternatives, and investments onto commit-ment (see Table 3). Consistent with Equation3, rewards and investments were positivelypredictive of commitment, whereas alterna-tives was negatively predictive of commit-ment. However, the costs factor was not sig-nificantly related to commitment (when theeffects on commitment of other variableswere simultaneously taken into account).3

Men versus women. The above-reportedanalyses were performed on selected subsam-ples (men vs. women, stayers vs. leavers, Stage1 vs. Stage 2 of involvement) to assess thegeneralizability of the model. The results ofthese analyses are summarized in Table 4.For both men and women, as for the sampleas a whole, greater rewards produced greatersatisfaction, but variations in costs did notsignificantly affect satisfaction. Greater sat-isfaction and investment size led to higherlevels of commitment for both men andwomen, but although low alternative qualityincreased commitment for women, alterna-tive quality did not significantly contributeto the prediction of commitment for men(this was true only for a model of commit-ment including satisfaction and investments

Note, REW = rewards, CST = costs, SAT = satisfaction,ALT = alternatives, INV = investments, COM = com-mitment, and SUBJ = subjects factor.;*p<.001.

31 speculated that the differential between a subject'sperceived rewards and costs might better predict satis-faction and commitment, so I computed an algebraicdifference score at each time for each subject and re-gressed this measure onto satisfaction, again using SUBJas a categorical factor. This multiple regression ac-counted for less of the variance in satisfaction (R2 = .58)than did the model including rewards and costs as twoseparate predictors. The same was true for the predictionof commitment; although the reward/cost, differentialmeasure was significantly correlated with commitment(j3 = .450), the multiple regression of reward/cost dif-ferential, alternatives, and investments onto commit-ment produced a lower /J2(.82) than did the model fromrewards, costs, alternatives, and investments.

INVESTMENT MODEL 109

as well as alternatives). The results for theprediction of commitment from a four-factormodel—rewards, costs, alternatives, and in-vestments—revealed that for men, greaterrewards and investments and poorer alter-natives increased commitment (as pre-dicted), and greater costs encouraged greatercommitment (contrary to predictions). Forwomen, greater rewards and investments andpoorer alternatives encouraged commitment,and costs did not significantly affect com-mitment (when the effects of other variablesare simultaneously taken into account).

Stayers versus leavers. The results forboth stayers and leavers were similar to thosefor the sample as a whole. As for the overallsample, for both stayers and leavers, rewardssignificantly affected satisfaction, but costsdid not. Also, for both stayers and leavers,commitment was increased by greater satis-faction, poorer alternatives, and larger in-vestment size. For the four-factor model of

commitment, greater rewards and invest-ments and poorer alternatives increased com-mitment. Costs increased commitment forstayers and did not significantly affect com-mitment for leavers.

Stage 1 versus Stage 2 The results re-ported above for the overall sample generallyhold true at all stages of involvement. Anal-yses comparing Stage 1 of relationships (Time0 to Time 5) to Stage 2 (Time 6 to Time 12)of relationships revealed that for both Stage1 and Stage 2, the three-factor model of com-mitment behaved as predicted, and the four-factor model showed that rewards, alterna-tives, and investments (but not costs) signif-icantly affected commitment as predicted(and in the same manner as for the overallsample). At Stage 1 satisfaction was signifi-cantly affected by rewards but not costs,whereas at Stage 2 greater rewards increasedsatisfaction and greater costs decreased sat-isfaction.

Table 4Multiple Regression Tests of Investment Model Predictions: Selected Subsamples

Model

MenSAT fromCOM fromCOM from

WomenSAT fromCOM fromCOM from

StayersSAT fromCOM fromCOM from

LeaversSAT fromCOM fromCOM from

Stage 1SAT fromCOM fromCOM from

Stage 2SAT fromCOM fromCOM from

+REW**+SAT**+REW**

+REW**+SAT**+REW**

+REW**+SAT**+REW**

+REW**-(-SAT**+REW**

+REW**+SAT**+REW**

+REW**+SAT**+REW**

CSTALT

+CST**

CST-ALT**

CST

CST-ALT**+CST*

CST-ALT**

CST

CST-ALT**

CST

-CST**-ALT**

CST

(SUBJ)+INV**-ALT**

(SUBJ)+INV**-ALT**

(SUBJ)+INV**-ALT**

(SUBJ)+INV**-ALT**

(SUBJ)+INV**-ALT*

(SUBJ)+INV*-ALT**

(SUBJ)+INV**

(SUBJ)+INV**

(SUBJ)+INV**

(SUBJ)+INV**

(SUBJ)+INV**

(SUBJ)+INV*

(SUBJ)

(SUBJ)

(SUBJ)

(SUBJ)

(SUBJ)

(SUBJ)

.90

.92

.93

.87

.88

.88

.90

.91

.91

.82

.83

.86

.91

.90

.92

.88

.93

.93

73.1485.9694.76

64.7865.7162.10

89.2395.6994.24

27.2026.1029.84

47.2438.1449.38

32.9262.3054.47

Note. REW = rewards, CST = costs, SAT = satisfaction, ALT = alternatives, INV = investments, COM = com-mitment, and SUBJ = subjects factor. + = a positive regression coefficient for a factor, and - = a negative regressioncoefficient.*p<.01.**p<.001.

110 CARYL E. RUSBULT

Table 5Multiple Regression Tests of Investment ModelPredictions: Overall Sample (Trend Scores)

Measure B R2 F

SAT fromREWCST

COM fromSATALTINV

COM fromREWCSTALTINV

.899**

.063

.244*-.459**

.397**

.238*

.043-.577**

.395**

.76 47.09

.78 34.26

.77 23.28

Note. REW = rewards, CST = costs, SAT = satisfaction,ALT = alternatives, INV = investments, and COM =commitment.*p<.05. **;><.01.

Processes of Change Over Time

Changes over time in investment modelvariables. To explore the manner in whichinvestment model factors change over time,each model variable was individually re-gressed onto time (with values 0 through 12),including SUBJ as a categorical variable.These analyses revealed that over time in re-lationships, rewards increased (/? = .402; t =6.93, p < .001), costs increased (/? = .316;t = 5.77, p < .001), satisfaction increased(0 = .189; t = 5.11, p < .001), alternativequality declined (/8 = -.179; t = -3.19, p <.002), investment size increased (0 = .772;t = 10.25, p < .001), and degree of commit-ment increased (|8 = .157; t = 6.57, p< .001).

Predicting satisfaction and commitment:Trend measures. To facilitate the remaininganalyses, I sought to develop a single measureof each investment model variable for eachsubject, measures reflecting the manner inwhich each variable changes over time,foreach subject. To do this, I performed analysesfor each of 34 subjects wherein each modelvariable was regressed onto time (0 through12). The resultant values, for each subject,represented the slope over time of changes ineach investment model variable. To assesswhether changes over time in satisfaction andcommitment result from correspondingchanges in investment model predictors, these"trend" scores were used in multiple regres-

sion analyses similar to those reported above(it was no longer necessary to include SUBJas a factor, because each subject now hadonly one set of measures). The results of theseanalyses, presented in Table 5, mirror thefindings reported above for the overall sam-ple. Satisfaction increased as a function ofincreases in obtained rewards (but not costs).The process of increasing satisfaction, de-creasing alternative quality, and increasinginvestment size produced corresponding in-creases over time in commitment. For thefour-factor model, increases over time incommitment were associated with increasingrewards, declining alternatives, and increas-ing investments but not by any significantchanges in costs.

Men versus women: Trend measures. Atwo-level multivariate analysis of variance(men, women) was performed on the trendscores to determine whether the process ofchange over time in investment model vari-ables differed as a function of sex. The mul-tivariate effect of sex on investment modelvariables was not significant, F(6, 26) = .77,p < .60, and univariate analyses revealed thatthe slopes over time for men and women ofrewards (F = .02), costs (F = .01), satisfaction(F = .02), alternatives F(= .25), investments(F = .21), and commitment (F = 1.79) didnot differ significantly.

Predicting Stay/Leave Behaviors

Stay/leave behaviors and investment modelfactors. A first step in the analysis of therelationship between investment model vari-ables and stay/leave behaviors was to estab-lish the simple effects of changes over timein model variables on behavior. To do thisa two-level multivariate analysis of variance(stayers, leavers) was performed on the trendscores. The results of this analysis are dis-played in Table 6. The multivariate effect ofstay/leave on investment model variables wassignificant, F(6, 26) = 5.33, p < .001, as wereall univariate effects (see Table 6). Comparedto leavers', stayers' rewards increased moreover time, costs increased less, satisfactionincreased more, alternative quality declinedmore (leavers' reported alternative qualityincreased somewhat), investment size in-creased more, and commitment increased

INVESTMENT MODEL in

Table 6Mean Trends Over Time in Investment Model Variables for Stayers and Leavers

Measure Stayers Abandoned Leavers 4f P<

REWCSTSATALTINVCOM

n

.488

.159

.423-.411

.542

.514

24

Two-level MANOVA

.145

.641

.117

.189

.155-.078

10.9810.994.41

13.056.34

18.91

313131313131

.002

.002,044.001.017.001

Three-level MANOVA

REWCSTSATALTINVCOM

n

.488

.159

.423-.411

.542

.514

24

.292

.458

.353-.154

.598

.248

4

.027

.779-.072

.463-.200-.339

5 •

6.886.403.89

10.1210.6315.32

2,302,302,302,302,302,30

.003

.005

.032

.001

.001

.001

Note, REW = rewards, CST = costs, SAT = satisfaction, ALT = alternatives, INV = investments, and COMcommitment. MANOVA = multivariate analysis of variance.

more (leavers' commitment declined overtime).

The leavers group, however, includes twotypes of individuals, those who decided toend their relationships and those whose part-ners decided to end their relationships. Be-cause the processes by which investmentmodel factors change over time may differ forthese two types of relationship, a third group,those who were "abandoned," was added tothe analysis reported above. The results ofthis analysis, too, are presented in Table 6.The multivariate effect of stay/abandoned/leave on investment model trend scores wassignificant, ^(12, 52) = 3.46, p < .001, aswere all univariate effects (see Table 6). Ofcourse, the stayers group's data performed inthe same manner as reported above (theirgroup status did not change in these analy-ses). The newly defined leavers group (thosewho personally ended their relationships)showed very little increase over time in re-wards, great increases in costs, a slight re-duction in satisfaction, an increase in alter-native quality, a decrease over time in in-vestments (divestiture), and a decline in levelof commitment. (No wonder they ended theirrelationships!) The abandoned group (thosewhose partners ended their relationships)

showed an interesting pattern of results:Compared to stayers', their rewards increasedless and costs increased more, so their sat-isfaction did not increase much over time.However, their alternatives continued to de-cline in quality and they continued to investas much in their relationships as did stayers.Thus, they reported moderately increasingcommitment and did not end their involve-ments. Collectively, these data suggest thatchanges over time in investment model vari-ables account for individuals' stay/leave be-haviors.4

Does commitment mediate stay/leave be-haviors? Ideally, it would be useful to usepath analysis techniques to determine whethercommitment mediates stay/leave behaviors(cf. Asher, 1976). Unfortunately, the stay/leave and stay/abandoned/leave measures arecategorical rather than continuous, so path

41 computed individual trend scores for the reward/cost differential measure and found that although thismeasure significantly distinguished between stayers andleavers—stayers' M = .276, leavers' M = -.244, F(\,31)= 7.72, p < .009; stayers' M = .276, abandonedM = -.121, leavers' M = -.342, F(2, 30) = 4.03, p <.028—the predictions from rewards and costs consideredindividually were superior.

112 CARYL E. RUSBULT

analysis procedures, which employ multipleregression analyses, are inappropriate. How-ever, we can use the general logic of path anal-ysis to perform related analyses and obtainsuggestive evidence regarding commitment'smediating role. The results of a series of mul-tiple regression analyses using the Equation2 and Equation 3 models of commitment andusing both stay/leave and stay/abandoned/

Table 7Predicting Stay/Leave Behaviors FromInvestment Model Factors

Table 7 (continued)

Behavior

Behavior /3

Prediction from

ST/LV fromSATALTINVCOM

ST/LV fromREWCSTALTINVCOM

S/A/L fromSATALTINVCOM

S/A/L fromREWCSTALTINVCOM

individual factors

-.511.544

-.412-.616

-.353.512.544

-.412-.616

-.430.624

-.560-.701

-.560.547.624

-.560-.701

Prediction from simple models

ST/LV fromSATALTINV

ST/LV fromREWCSTALTINV

S/A/L fromSATALTINV

S/A/L fromREWCSTALTINV

-.343.638

-.288

-.132.226.275

-.168

-.266.713

-.405

-.068.154.387

-.338

Prediction including commitment

ST/LV fromSATALTINVCOM

ST/LV fromREWCSTALTINVCOM

S/A/L fromSATALTINVCOM

S/A/L fromREWCSTALTINVCOM

-.326.461

-.061-.528

-.078.243.051

-.014-.388

-.363.531

-.247-.498

-.029.166.222

-.225-.385

Note. REW = rewards, CST = costs, SAT = satisfaction,ALT = alternatives, INV = investments, COM = com-mitment, ST/LV = stay/leave, and S/A/L = stay/aban-doned/leave.

leave as criteria are summarized in Table 7.The regression of commitment onto bothcriteria yielded fairly sizable betas (-.616 and-.701, respectively). Regressing commit-ment predictors individually onto both cri-teria revealed impressive betas for both Equa-tion 2 and Equation 3 models (see Table 7,"Prediction from individual factors"). Whenthe commitment predictors—Equations 2and 3—were collectively regressed onto stay/leave and stay/abandoned/leave, their regres-sion coefficients (generally) declined some-what (see Table 7, "Prediction from simplemodels"). When the commitment measurewas added to these simpler models, theregression coefficients for the commitmentpredictors (generally) declined even further,whereas the commitment coefficients re-mained relatively strong (see Table 7, "Pre-diction including commitment"). Thus,changes over time in commitment were morestrongly predictive of stay/leave and stay/abandoned/leave than were other investmentmodel factors, and although they (SAT, ALT,INV or REW, CST, ALT, INV) collectively

INVESTMENT MODEL 113

provided a reasonable prediction of both cri-teria, when the impact of commitmentchanges on stay/leave behaviors was ac-counted for, their impact declined. Thesefindings suggest that although rewards, costs,satisfaction, investments, and alternativesmay exert some relatively small direct effecton stay/leave behaviors, their impact on stay/leave is largely indirect, mediated by changesover time in commitment.

Intent commitment versus attachmentcommitment. The investment model assertsthat commitment consists of two compo-nents—behavioral intent and psychologicalattachment. The above-reported analysescombined these two. components, studyingcommitment as a single construct. Is this ap-proach reasonable? Do these two compo-nents behave similarly? To answer this ques-tion I divided the five measures of commit-ment into two separate (averaged) subscales:intent ("How likely is it that you will endyour relationship in the near future?"; "Forwhat length of time would you like your re-lationship to last?"; and "How attractive analternative would you require before adopt-ing it and ending your relationship?") andattachment ("To what extent are you 'at-tached' to your partner?" and "To what ex-tent are you committed to your relation-ship?"). I regressed the attachment measureonto the intent measure, including SUBJ asa categorical factor, and found that these twomeasures were strongly related (R2 = .76,(3 = .456). I also computed individual trendscores for each subscale and found thatchanges over time in intent and attachmentwere significantly correlated (r = .82). Fur-thermore, changes over time in both intentand attachment distinguished between stayersand leavers—for intent, stayers' M = .372,leavers' M = -.449, F(\, 31) = 19.05, p <.001; for attachment, stayers' M = .489, leav-ers' M= -.075,^(1, 31) = 11.62,p<.002—and among stayers, those who were aban-doned, and leavers—for intent, stayers' M =.371, abandoned M = -.093, leavers' M =-.734, F(2, 30) = 12.74, p < .001; for at-tachment, stayers' M = .489, abandonedM= .327, leavers' M = -.396, F(2, 30) =11.08, p < .001. These results suggest thatbehavioral intent artd feelings of psycholog-ical attachment are strongly related to one

another and that (at least in the present in-vestigation) it was reasonable to deal withthese two aspects of commitment as a singletheoretical construct.

Discussion

The results of the present study providegenerally good support for investment modelpredictions regarding the development anddeterioration of satisfaction and commit-ment and the causes of individuals' stay/leavebehaviors. The investment model assertionconcerning the impact of rewards and costson satisfaction was assessed first (Equation1). Increases in rewards consistently led togreater satisfaction, whereas variations incosts did not significantly affect level of sat-isfaction. This pattern of results held true notonly for the sample as a whole but also forboth men and women, for both stayers andleavers, and for early stages of involvement.The only exception to this overall pattern wasfor later stages of involvement (from about3 to 7 months of involvement), at which timeincreases in costs led to significant decreasesin level of reported satisfaction. This patternof results may have occurred because thecosts of a relationship are simply not appar-ent at early stages of involvement. At thebeginning of a relationship, individuals maytry hard to display their best selves, and theirpartners may be generous in overlooking anyfaults (or problems) that do become appar-ent. However, at later stages of involvement,persons may relax more, allow their true,flawed selves to emerge, and their partnersmay adopt a more realistic view of the re-lationship.

A second task was to explore the ability ofinvestment model variables to predict levelof commitment to maintain relationships.The three-factor model expressed in Equa-tion 2—from satisfaction, alternatives, andinvestments—behaved fairly consistently aspredicted. Greater satisfaction and invest-ment size and poorer alternatives promotedhigher levels of commitment for the overallsample, for both stayers and leavers, at allstages of involvement, and for women. Formen the impact of alternatives on commit-ment was not significant (at least not for thisthree-factor model). (Because poorer alter-

114 CARYL E. RUSBULT

natives did promote greater commitment formen in the four-factor model, this finding wasnot of great concern; perhaps their reportedsatisfaction was simply highly colinear withreported alternatives.) In regard to the four-factor model of commitment (Equation 3)—from rewards, costs, alternatives, and invest-ments—it was found that greater rewards,poorer alternatives, and greater investmentsize encourage higher levels of commitment,whereas variations in costs generally have nosignificant effect on commitment. These find-ings are consistent with my earlier results(Rusbult, 1980a), arguing that "the romanticideal that one accepts a mate 'for better orworse' may prevent individuals from admit-ting that they become less committed to oneanother as the costs of doing so increase" (p.184). The only exceptions to this finding werefor men and stayers, for whom greater costsactually increased level of commitment. Thismay be because a third variable—time—leads to increases in both costs and commit-ment, or this phenomenon may result froma sort of dissonance or self-perception effect(e.g., "There are numerous costs in this re-lationship, but I remain involved, so I mustbe really committed."). That later analysesof trends over time in costs and commitmentrevealed no significant relation suggests thatthe former explanation is more probable (i.e.,when the effects of time are accounted for,this relation is not apparent). Alternatively,for some persons at some times, costs mayserve as a sort of investment (i.e., sunk costs);previous costs incurred with little rewardmay lead to increased determination to makethose costs "pay off" in the future. This eco-nomic principle—that people are tempted to"throw good money after bad money"—isevident in research on social behavior in non-romantic settings (e.g., Rubin & Brockner,1975; Staw, 1976; Staw & Fox, 1977) andmay account for some of the aberrant find-ings in the present study. This speculationremains to be explored.

The study also sought to explore changesover time in investment model factors. Overtime, rewards increase, costs increase, satis-faction increases, alternative quality declines,investment size increases, and level of com-mitment increases. It seems natural that in-vestment size increases over time because a

variety of resources require time for their in-vestment. Also, it is gratifying to discover thatlevel of rewards, satisfaction, and commit-,ment becomes greater over time. Unfortu-nately, costs too increase over time. As men-tioned earlier, perhaps early stages of involve-ment are characterized by infatuation, inwhich both partners are "on their best be-havior" and are generous in their evaluationsof one another (i.e., they perceive fewercosts); we may allow our less attractive traitsto become evident only at later stages of in-volvement. It is somewhat surprising to notethat individuals' perceptions of their alter-natives decline over time. This seems reason-able because alternatives affect commitmentand commitment increases over time, butthere is no reason to assume from investmentmodel assertions that within a given individ-ual, alternative quality should decline overtime. This phenomenon may occur becauseindividuals' alternatives really do declineover time in an involvement (e.g., alternativepartners do not approach them because ofawareness of their involvement) or becausepersons simply come to perceive their alter-natives as less attractive with the passage oftime and increased involvement.

An analysis of trend scores (i.e., measuresreflecting changes over time in each modelvariable for each subject) revealed a patternof findings identical to that summarized ear-lier. Increases over time in rewards led to cor-responding increases in satisfaction, butchanges over time in costs did not signifi-cantly affect satisfaction. Increases over timein commitment appear to result from in-creases in satisfaction, declines in the qualityof available alternatives, and increases in in-vestment size. Tests of the Equation 3 modelof commitment revealed that commitmentincreases over time because of correspondingincreases in rewards, declines in alternativequality, and increased investment. Variationsin level of costs over time did not significantlyaffect commitment. ,

A final goal, perhaps the most importantgoal of the present study, was to determinethe relations between investment model vari-ables and stay/leave behaviors. Two types ofanalysis—for stayers versus leavers and forstayers versus abandoned individuals versusleavers—revealed similar patterns of find-

INVESTMENT MODEL 115

ings. For stayers, rewards increased over time,costs rose slightly, satisfaction grew, alterna-tive quality declined, investment size in-creased, and level of commitment grew. Forleavers, rewards did not increase very much,costs increased greatly, satisfaction declinedsomewhat, alternative quality greatly im-proved, investment size declined, and com-mitment declined. It is interesting to notethat some form of "divestiture" occurs amongleavers; over time, they seem to remove (orreclaim) resources invested in their relation-ships at earlier times. Individuals who wereabandoned evidenced an intriguing patternof change over time, a pattern of change thatcould be termed entrapment. These personsshowed fewer increases in rewards, greaterincreases in costs, and lower increases in levelof satisfaction than did stayers (i.e., they werenot tremendously happy with their involve-ments). However, their alternatives declinedin quality and they continued to investheavily in their relationships (this group in-vested at as great a rate, or greater, than didstayers). They therefore reported moderatelevels of commitment and remained involved(albeit trapped) in their relationships untilthey were terminated by their partners.

A final issue addressed in this study wasthe role of variations in commitment in me-diating stay/leave behaviors (Equation 4). Inall cases—examining stay/leave or stay/aban-doned/leave criteria, using three- or four-fac-tor models of commitment—commitment isa better zero-order predictor of stay/leavebehaviors than is any other investment modelvariable. Moreover, when simple models ofcommitment (satisfaction, alternatives, andinvestments or rewards, costs, alternatives,and investments) as predictors of stay/leaveare compared to similar models, includingcommitment as a predictor of stay/leave, thevariance in stay/leave accounted for by com-mitment predictors declines, whereas com-mitment itself continues to exert a sizableinfluence on stay/leave behaviors. Althoughthese conclusions should be regarded as ten-tative (the analyses were not entirely "ko-sher" in regard to the assumptions of regres-sion-type analyses), these results are consis-tent with investment model predictions.

Thus, the present study reveals fairly goodsupport for investment model hypotheses.

These findings provide interesting evidenceregarding the process by which satisfactionand commitment develop and deteriorateover time. The study also provides suggestiveevidence concerning the role of commitmentin mediating individuals' stay/leave behav-iors. However, several limitations of this in-vestigation should be noted. First, althoughthe attempt to study 34 college students' dat-ing relationships over a 7-month time periodmay be regarded as an ambitious endeavor,one may nevertheless question the general-izability of these findings—the findings arebased on a small number of very youngadults involved in the early stages of relativelyshort-term involvements. Three of our 34subjects later married their partners, but itwould be useful to conduct a similar studyusing an older subject population with morelong-standing involvements.

Second, it should be noted that this studyexamined self-reported satisfaction and com-mitment and self-reported rewards, costs, al-ternatives, and investments. It is encouragingto find that these factors effectively predictstay/leave behaviors, but it is still unclearwhat relation self-report has to actual, ob-jective conditions (e.g., Do reported rewardsaccurately represent real, obtained rewards?).

Third, a related issue concerns subjects'ability to make dependable distinctionsamong abstract concepts (e.g., Can subjectsaccurately distinguish between rewards andsatisfaction, between costs and investmentsize? Do subjects strive for consistency acrossmeasures? Do they answer based on a simplehalo effect?). All of the above-reported anal-yses in some manner controlled for thewithin-subjects factor, either by includingsubject as a factor or by eliminating individ-ual response tendencies through the com-putation of individual trend scores. Such pro-cedures should, at least in part, eliminateproblems associated with consistency tenden-cies or halo effects. Also, the format of thequestionnaires was designed to help subjectsdistinguish among concepts—definitions ofall variables were presented, and both con-crete and general questions were answered.In addition, the findings themselves suggestthat subjects were able to distinguish amongconcepts. For example, subjects were able toreport that both costs and satisfaction in-

116 CARYL E. RUSBULT

creased over time; abandoned persons re-ported that in spite of increasing costs, theycontinued to invest in their relationships.

Finally, one must question the effects thatthe repeated administration of question-naires has on subjects' responses: Do subjectsstrive for consistency over time, does thisprocedure produce reactive effects, do sub-jects develop self-fulfilling expectations, andso on? The inclusion of the within-subjectsfactor in the analyses presumably controls formany statistical problems associated with re-peated administration. That similar patternsof results were obtained during both earlyand later stages of involvement suggests thatreactive effects were not a serious problem.Information on nonparticipant volunteersfor the study was also obtained to determinewhether participation affected stay/leave de-cisions. Thirty-one percent of the nonparti-cipants' relationships ended during the 7-month period of the study, and 29% of thesubjects' relationships ended. Thus, the actof completing a questionnaire once every 17days does not seem to have significantly in-fluenced stay/leave behaviors. Thus, althoughlongitudinal procedures introduce a numberof important methodological difficulties, thebenefits gained through the conduct of thisstudy using a longitudinal method by far out-weigh the problems associated with suchmethods.

Although the research reported herein usesthe investment model as a theoretical tool forexploring longitudinal effects in romantic re-lationships, I do not wish to argue that thismodel should replace traditional researchand theory on romantic involvements. Theinvestment model may be a useful means ofintegrating diverse findings regarding the roleof variables such as attitudinal similarity,physical attractiveness, and self-disclosures,but it is still important that such phenomenabe explored in greater detail on their own.For example, the results of this study arefairly consistent with those of Hill et al.'s(1976) 2-year longitudinal study of breakupsbefore marriage. These authors found thatfactors such as liking and loving (i.e., satis-faction), estimates of marriage probability(i.e., commitment), duration and exclusivity(i.e., investments), and partner's similarity(i.e., rewards, costs) effectively predicted

breakups. Although the present approach ismore strongly grounded in the extant theoryon relationships, their study explores ingreater detail factors such as the timing ofbreakups and partner's explanations forbreakups. Also, although these data were ex-amined from the point of view of the in-vestment model, these findings are not in-consistent with theories such as social-pene-tration theory (Altman & Taylor, 1973) orincremental-exchange theory (Huesmann &Levinger, 1976). For example, Levinger (1979)distinguishes between attractive versus un-attractive and stable versus unstable relation-ships. The analyses of abandoned individualsin the present study reveal a pattern remi-niscent of his "unattractive stability" group,except that such conditions appear to resultnot only from low attraction and poor qualityalternatives but also from high levels of in-vestment. The utility of the investment modelmay lie in its applicability to a wide range ofsocial-exchange relations (friendships, ro-mantic involvements, and job-related behav-iors; Farrell & Rusbult, 1981; Rusbult, 1980a,1980b; Rusbult & Farrell, in press), in itsability to integrate the diverse literature onromantic relationships, and in its foundationin more general theories of social relations,particularly interdependence theory (Kelley& Thibaut, 1978; Thibaut & Kelley, 1959).It is hoped that through such functions theinvestment model may contribute to our un-derstanding of social behavior.

References

Ajzen, I. Effects of information on interpersonal attrac-tion: Similarity versus affective value. Journal of Per-sonality and Social Psychology, 1974, 29, 374-380.

Altman, I., & Taylor, D. A. Social penetration: The de-velopment of interpersonal relationships. New York:Holt, Rinehart & Winston, 1973.

Asher, H. B. Causal modeling. Beverly Hills, Calif.: Sage,1976.

Becker, H. S. Notes on the concept of commitment.American Journal of Sociology, 1960, 66, 32-40.

Bentler, P. M., & Newcomb, M. S. Longitudinal studyof marital success and failure. Journal of Consultingand Clinical Psychology, 1978,47, 1053-1070.

Berscheid, E., Brothen, X, & Graziano, W. Gain-losstheory and the "law of infidelity": Mr. Doting versusthe admiring stranger. Journal of Personality and So-cial Psychology, 1916,33, 709-718.

Billings, A. Conflict resolution in distressed and nondis-tressed married couples. Journal of Consulting andClinical Psychology, 1979, 47, 368-376.

INVESTMENT MODEL 117

Birchler, G. R., Weiss, R. L.,, & Vincent, J. P. Multi-method analysis of social reinforcement exchange be-tween maritally distressed and nondistressed spouseand stranger dyads. Journal of Personality and SocialPsychology, 1975, 31, 349-360.

Blau, P. M. Exchange and power in social life. New York:Wiley, 1964.

Brown, P., & Manela, R. Changing family roles: Womenand divorce. Journal of Divorce, 1978, 1, 315-328.

Cohen, J., & Cohen, P. Applied multiple regression/cor-relation analysis for the behavioral sciences. Hillsdale,N.J.: Erlbaum, 1975.

Drachman, D., DeCarufel, A., & Insko, C. A. The extracredit effect in interpersonal attraction. Journal ofExperimental Social Psychology, 1978, 14, 458-465.

Farrell, D., & Rusbult, C. E. Exchange variables as pre-dictors of job satisfaction, job commitment, and turn-over: The impact of rewards, costs, alternatives, andinvestments. Organizational Behavior, and HumanPerformance, 1981, 27, 78-95.

Fineberg, B. L., & Lowman, J. Affect and status dimen-sions of marital adjustment. Journal of Marriage andthe Family, 1975, 37, 155-160.

Gottman, J,, et al. Behavior exchange theory and maritaldecision making. Journal of Personality and SocialPsychology, 1976, 34, 14-23.

Hatfield, E., Utne, M. K., & Traupman, J. Equity theoryand intimate relationships. In R. L. Burgess & T. L.Huston (Eds.), Social exchange in developing relation-ships. New York: Academic Press, 1979.

Hill, C. T, Rubin, Z., & Peplau, L. A. Breakups beforemarriage: The end of 103 affairs. Journal of SocialIssues, 1976, 32, 147-168.

Homans, G. C. Social behavior: Its elementary forms.New York: Harcourt, Brace & World, 1961.

Huesmann, L. R., & Levinger, G. Incremental exchangetheory: A formal model for progression in dyadic so-cial interaction. In L. Berkowitz & E. Walster (Eds.),Advances in experimental social psychology (Vol. 9).New York: Academic Press, 1976.

Insko, C. A., & Wilson, M. Interpersonal attraction asa function of social interaction. Journal of Personalityand Social Psychology, 1977, 35, 903-911.

Johnson, M. P. Commitment: A conceptual structureand empirical application. Sociological Quarterly,1973,14, 395-406.

Kelley, H. H., & Thibaut, J. W. Interpersonal relations:A theory of interdependence. New York: Wiley, 1978.

Kerckhoff, A. C. Patterns of marriage and family for-mation and dissolution. Journal of Consumer Re-search, 1976, 2, 261-275.

LaGaipa, J. J. Interpersonal attraction and social ex-change. In S. Duck (Ed.), Theory and practice in in-terpersonal attraction. New York: Academic Press,1977.

Levinger, G. A three-level approach to attraction: Towardan understanding of pair relatedness. In T. L. Huston(Ed.), Foundations of interpersonal attraction. NewYork: Academic Press, 1974.

Levinger, G. A social exchange view on the dissolutionof pair relationships. In R. L. Burgess & T. L. Huston(Eds.), Social exchange in developing relationships.New York: Academic Press, 1979.

Levinger, G., & Moles, O. C (Eds.). Divorce and sepa-ration. New York: Basic Books, 1979.

Nunnally, J. Psychometric theory. New York: McGraw-Hill, 1967.

Rosenblatt, P. C. Needed research on commitment inmarriage. In G. Levinger & H. L. Raush (Eds.), Closerelationships: Perspectives on the meaning of intimacy.Amherst: University of Massachusetts Press, 1977.

Rubin, J. Z., & Brockner, J. Factors affecting entrapmentin waiting situations: The Rozencrantz and Guilden-stern effect. Journal of Personality and Social Psy-chology, 1975, 31, 1054-1063.

Rusbult, C. E. Commitment and satisfaction in romanticassociations: A "test of the investment model. Journalof Experimental Social Psychology, 1980, 16, 172-186. (a)

Rusbult, C. E. Satisfaction and commitment in friend-ships. Representative Research in Social Psychology,1980, 11, 96-105. (b)

Rusbult, C. E., & Farrell, D. A longitudinal test of theinvestment model: The impact on job satisfaction, jobcommitment, and turnover of variations in rewards,costs, alternatives, and investments. Journal of AppliedPsychology, in press.

Staw, B. M. Knee-deep in the big muddy: A study ofescalating commitment to a chosen course of action.Organizational Behavior and Human Performance,1976, 16, 27-44.

Staw, B. M., & Fox, F. V. Escalation: The determinantsof commitment to a chosen course of action. HumanRelations, 1977, 30, 431-450.

Thibaut, J. W., & Kelley, H. H. The social psychologyof groups. New York: Wiley, 1959.

Todres, R. Runaway wives: An increasing North Amer-ican phenomenon. Family Coordinator, 1978,27, 17-21.

Tyler, T. R., & Sears, D. O. Coming to like obnoxiouspeople when we must live with them. Journal of Per-sonality and Social Psychology, 1977, 35, 200-211,

Walster, E., Walster, G. W, Piliavin, J., & Schmidt, L."Playing hard to get"; Understanding an elusive phe-nomenon. Journal of Personality and Social Psychol-ogy, 1973,26, 113-121.

Received August 16, 1982Revision received November 15, 1982