Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2012 May 9.
Published in final edited form as: J Fam Issues. 2010 Oct 18;32(6):820–841. doi: 10.1177/0192513X10385788

The Revised Commitment Inventory: Psychometrics and Use with Unmarried Couples

Jesse Owen 1, Galena K Rhoades 2, Scott M Stanley 2, Howard J Markman 2
PMCID: PMC3348588  NIHMSID: NIHMS373434  PMID: 22582000

Abstract

The Commitment Inventory (Stanley & Markman, 1992) measures interpersonal commitment (dedication) and constraint commitment. Since it was first published, substantial revisions have been made, but there are no published data on the psychometric properties of the new version. Further, little information is available on measuring commitment for unmarried couples. This study examined the psychometric properties of the Commitment Inventory in 320 premarital or cohabiting couples (N = 640). Dyadic confirmatory factor analyses revealed that the hypothesized factor structure of six constraint subscales and one dedication subscale fit the data well for both men and women. Internal consistency coefficients were within acceptable range for most subscales. Within-couple correlations as well as correlations among subscales and with relationship quality, negative communication, and religiosity are presented. Implications for future research are discussed.

Keywords: Commitment, Romantic Relationships, Psychometric, Confirmatory Factor Analysis, Couples

Although it represents a relatively new area of research, commitment is an important facet of romantic relationships and their development. Various aspects of commitment have now been shown to be associated with risk factors as well as protective qualities in relationships (e.g., Adams & Jones, 1997; Impett, Beals, & Peplau, 2002; Rusbult, 1983; Stanley, Whitton, & Markman, 2004; Stanley, Whitton, Sadberry, Clements, & Markman, 2006; Wiesequist, Rusbult, Foster, & Agnew, 1999).

Since social science advances by both adequate theoretical specification and measurement of important constructs, measurement of commitment remains an important matter in this burgeoning field. Over a decade ago, Adams and Jones (1997), in a review of measures of commitment, noted that “[t]here were relatively few studies that addressed construct validity directly (for an exception, see Stanley & Markman, 1992) (p. 1178).” Since that time, Stanley and Markman's (1992) Commitment Inventory (in whole or parts) has been used in a variety of studies examining romantic relationships (e.g., Pistole & Vocaturo, 1999; Demoss, 2004; Spiers, 1998), especially studies using the dedication construct and subscale (e.g., G. Kline et al., 2004; Stanley, Amato, Johnson, & Markman, 2006; Whitton, Stanley, & Markman, 2006). Despite this activity, there has not been an investigation of the Commitment Inventory's factor structure since the original 1992 article, even though several new subscales have been added and the format of some subscales has been altered. Moreover, there are no studies examining the factor structure of the Commitment Inventory with unmarried couples. Thus, the aims for the current study were to examine the psychometric properties of the Commitment Inventory in two large samples of unmarried couples.

Theoretical Underpinnings

Though many have written about what commitment is and how it relates to other aspects of romantic relationships (e.g., Adams & Jones, 1997; Agnew, Van Lange, Rusbult, & Langston, 1998; Ferguson, 1993; Johnson, Caughlin, & Huston, 1999; Rusbult et al., 1994; Stanley & Markman, 1992), there is no consensus on a definition. Most theories of commitment are based in part on interdependence theory which suggests that over time, couples' interdependence is formed through pro-relationship behaviors, such as forms of dedication (i.e., long-term view of the relationship, willingness to sacrifice) and their perceived loss if the relationship were to end (i.e., constraints) (Rusbult & VanLange, 2003). In turn, the degree of interdependence within a couple can influence each other's behaviors and perceptions of the relationship (e.g., Wieselquist et al., 1999). Specifically, Stanley and Markman's (1992) Commitment Inventory was based on a theory of commitment that involves these two main components and refers to them as dedication and constraint commitment.

Dedication describes the intrinsic desire to be with one's partner. It encompasses couple identity (e.g., “we are a team”), having a long-term view of the relationship, making the relationship a priority, and making sacrifices for one's partner or the good of the relationship. On the other hand, constraint commitment refers to aspects of one's relationship or partner that make it difficult to break up (e.g., financial investments, social pressure to stay together, concern for partner's welfare, and worry that there are no other suitable partners). Constraint commitment may seem negative or detrimental and it likely is in some cases, but it can also motivate partners to stay together and work through difficult times. However, constraint commitment alone is not sufficient to maintain a healthy relationship, and it has historically been far less associated with relationship quality than dedication is (Adams & Jones, 1997; Johnson et al., 1999; Stanley & Markman, 1992).

Previous Research and New Advances to Theory and Measurement of the Commitment Inventory

The initial version of the Commitment Inventory had 36 dedication items representing six subscales and 69 constraint items representing six subscales (Stanley & Markman, 1992). These subscales demonstrated good internal consistency, with Cronbach's alphas ranging from .70 to .94. The measure also was shown to be valid, as the constraint and dedication subscales correlated as predicted with other commitment and marital satisfaction measures (see Johnson et al., 1999; Stanley & Markman, 1992).

Stanley identified three weaknesses of the Commitment Inventory and subsequently revised the measure in the mid-1990s for use in a premarital education outcome study (see Stanley et al., 2001). First, the original version (as published in 1992) included two constructs, based on Johnson et al.'s (1999) work, that were measured in a format that was very different from that of other subscales: termination procedures (i.e., the difficulty of the steps needed to end a relationship) and overall alternative quality (i.e., the effect relationship dissolution would have on a person's overall well-being and lifestyle). These subscales were revised to be shorter and so that they followed the same response format as other items.

Second, the original version of the Commitment Inventory did not assess some types of constraint commitment that seemed important based on other research (Johnson, 1973; Johnson et al., 1999) and on clinical experience with couples. These new constructs encompassed concern for the welfare of children, concern for the welfare of one's partner, and what one's financial status would be if they relationship ended. These three constructs represent psychological and tangible reasons for staying with a partner and were not included in the original Commitment Inventory.

Third, the length of the original Commitment Inventory (105 items) was a concern. When they are psychometrically sound, brief measures tend to be advantageous in terms of time and cost, especially in studies that measure many constructs. As such, a short form for of the dedication construct was also developed. It collapses across the original six subscales to assess dedication with 14 items. This decision was in keeping with Stanley and Markman's (1992) earlier recommendation that dedication be assessed globally except when there are specific questions about its subdimensions. This shorter form has not been empirically vetted for its factor structure or reliability.

The current study tests the factor structure of this revised Commitment Inventory, as well as its reliability and validity. We focus on testing the psychometric properties of the Revised Commitment Inventory among cohabiting or premarital couples for several reasons. First, although commitment dynamics are likely particularly important in these relationship stages, they have rarely been examined and little is known about the utility of measures of commitment, particularly the Commitment Inventory, for these populations. Second, including cohabiting couples who may or may not have plans for marriage increases the heterogeneity of the sample, which increases variance and statistical power as well as the generalizability of findings.

In addition, we examine whether the factor structure is similar for men and women. The original version reportedly had similar structure across gender (Stanley & Markman, 1992), but there are a few reasons to suspect that items regarding commitment could be interpreted differently by men and women, especially in early stages of relationships (see Stanley, 2002). For instance, there is some evidence that men report lower levels of dedication than women in cohabiting relationships (Stanley et al., 2004) and in marriages that were preceded by cohabitation (Rhoades, Stanley, & Markman, 2006), suggesting that men and women could have different understandings of commitment, especially in unmarried unions. For comparisons between men and women's levels of commitment to be valid, the factor structure of the measure used must be similar for men and women. Thus, we tested for gender differences in the current study.

Hypotheses

First, we hypothesized that the Revised Commitment Inventory would be best represented by one dedication subscale and six separate constraint subscales, supporting Stanley and Markman's (1992) commitment theory (Hypothesis 1). To test this hypothesis, we utilized dyadic confirmatory analysis, which is necessary to effectively model the systematic error that arises from assessing men and women of the same couple. Moreover, it allowed us to test our second hypothesis that the factor structure would fit similarly for men and women (Hypothesis 2). We also present the internal consistency of the subscales using Cronbach's alpha coefficients. To assess criterion-related validity, the relationships between the Revised Commitment Inventory subscales and relationship adjustment, communication, and religiosity were tested. Based on previous research and on commitment theory (e.g., Stanley et al., 2001; Stanley & Markman, 1992), it was expected that the dedication subscale would be moderately correlated with relationship adjustment, negative communication, and religiosity, and that there would be only small correlations between constraint subscales and these variables (Hypothesis 3). The correlations between partners' scores on the Revised Commitment Inventory subscales were examined to test the hypothesis that within-couple correlations would be small to medium for matching subscales (Hypothesis 4).

Method

Participants for this study came from two larger studies. The samples are therefore described separately below.

Engaged Sample

The first sample was from an ongoing project on the effectiveness of relationship education that is detailed elsewhere (see Markman et al., 2004; Stanley et al., 2001). As part of the larger study, 306 heterosexual couples were recruited from the religious organizations where they were planning to marry. They completed an assessment before they received premarital training and approximately six weeks following their training (post). Here, only post data were used because this was the first assessment at which the full Revised Commitment Inventory was given. During the post assessment, partners completed questionnaires individually on computer or by paper and pencil and engaged in two videotaped discussions (not analyzed here). Couples were paid $40. Of the 242 couples who completed post, there were 42 couples in which one or both partners did not provide complete data on the Revised Commitment Inventory, thus, they were excluded from the present study. The final sample was 200 couples (N = 400).1

Overall, the participants were 27.36 years old (men's M = 28.07 SD = 5.70; women's M = 26.66 SD = 5.78), with a median education level of 16 years, and a median personal income level of $20,000–29,999 annually. The men were 84.5% White, 3.9% African American, 10.1% Hispanic or Latino(a), .5% Asian, and 1% Native American and women were 84.1% White, 2.9% African American, 10.1% Hispanic or Latino(a), and 2.4% Asian;.5% of women did not specify ethnicity. Over 95% of men and women had never been married before and approximately 14% of the couples had children (either from the current relationship or a past one). Couples had been dating, on average, 36.00 months (SD = 27.07) and 71% were cohabiting.

Cohabiting Sample

The participants for the second sample were from a longitudinal study on cohabiting relationships that is detailed by Rhoades, Stanley, and Markman (2009). Briefly, 120 couples in opposite-sex cohabiting relationships were recruited nationwide through online and printed announcements in newsletters or listservs. Couples who expressed interest received two sets of questionnaires (one for each partner), along with a cover letter that explicitly asked that partners not share answers. Partners mailed their forms back separately in postage-paid envelopes and were compensated by being entered into a $50 lottery. Data were collected two more times, four months apart, but only data from the first wave of the study were used in the present study.2 The response rate for the first wave (both partners returning forms) was 47.6%.

Overall, the participants were 28 years old (women's M = 27.74, SD = 5.69, men's M = 29.93, SD = 6.93). On average, they had completed 16 years of education and made $25,000 annually. The women in this sample were 82.5% White, 4.2% Asian, 4.2% Hispanic, .8% Black, and 4.1% other; 4.2% did not report their ethnicity. The men in this sample were 89.2% White, .8 Asian, 5.0% Hispanic, .8% Black, and 1.7% other; 2.5% did not report their ethnicity. The vast majority of women (89.17%) and men (87.4%) had never been married; 9.16% of couples reported that they had children living in their homes. Thirty couples (25%) were engaged. Couples had been dating for 41.21 months, on average (SD = 26.81 months).

Measures

Except where noted, both samples completed the same measures.

Commitment Inventory

As described earlier, the original Commitment Inventory included 105 items, but it was later revised to 55 items. In the current study, subscales that would not apply to unmarried couples were excluded, leaving a total of 36 items. Specifically, the subscale Concern for Children's Welfare was excluded because very few couples in the current study had children together. Further, we excluded the Morality of Divorce subscale since the couples were not currently married. This Revised Commitment Inventory was administered to both samples and included seven subscales. The Dedication subscale (14 items) refers to the priority of relationship, couple identity, satisfaction with sacrifice, and having a long-term view of the relationship. The Structural Investments subscale (4 items) assesses tangible assets (e.g., money) that are invested in the relationship. The Social Pressure subscale (5 items) refers to the pressures that family and friends assert on the couple to stay in the relationship. The Termination Procedures subscale (3 items) assesses the perception about how difficult it would be to end the relationship. The Concern for Partner Welfare subscale (3 items) refers to the belief about effects that ending the relationship would have on the partner's well-being. The Alternative Financial Status subscale (3 items) describes the degree to which a person's financial status would change if the relationship were to end. Lastly, the Alternative Availability subscale (4 items) describes the potential for dating other suitable partners if the current relationship ended. Detailed results (i.e., factor loadings, internal consistency, means, and standard deviations) are presented in Results.

Relationship adjustment

The two samples completed different measures of relationship adjustment. The engaged sample completed the Marital Adjustment Test (MAT; Locke & Wallace, 1959), a widely used measure of relationship quality with acceptable validity as well as the ability to discriminate between distressed and nondistressed couples (Crane, Allgood, Larson, & Griffin, 1990). Although this measure has historically demonstrated high levels of reliability, Cronbach's alphas (αs) were lower in this sample (i.e., .57 men and .65 for women). Sample characteristics probably constrained the reliability estimates (i.e., there were ceiling effects for the individual items in this study of relatively happy couples). Scores ranged from 57 to 158 (M = 128.55, SD = 16.10).

The cohabiting sample completed the seven-item version of the Dyadic Adjustment Scale (DAS), a measure of global relationship adjustment with high reliability and validity (Hunsley, Best, Lefebvre, & Vito, 2001; Spanier, 1976). Here, αs were .74 for men and .71 for women.

Religiosity

Religiosity was assessed by the question: “All things considered how religious would say you are?” This item has been shown to be a viable measurement for global religiosity (e.g., Rhoades et al., 2006; Owen, Rhoades, Stanley, & Fincham, in press).

Negative communication

The 8-item Communication Danger Signs Scale (Stanley & Markman, 1997) was used to assess dimensions of negative communication, including escalation, invalidation, and withdrawal. Respondents rate each item on a 1 (almost never) to 3 (frequently) scale. In a variety of samples of married and/or cohabiting couples or individuals, the measure has demonstrated high reliability and validity (e.g., Stanley, Markman, & Whitton, 2002). Cronbach's alpha for men and for women was .73 in the cohabiting sample and .80 for men and .74 for women in the engaged sample.

Data Analytic Plan

We first sought to test the factor structure of the Revised Commitment Inventory, with six constraint subscales (social pressure, available alternative partners, alternative financial status, concern for partner welfare, termination procedures, and structural investments) and one dedication subscale. Confirmatory factor analysis is typically indicated for later stages of scale development, particularly after a clear theory guiding the scale's purpose is delineated and after the scale has demonstrated some initial validity (R. Kline, 1998). Commitment theory is clear and initial validation of the Commitment Inventory has been conducted (Stanley & Markman, 1992), so confirmatory factor analysis is the most appropriate analysis for the current study; it allows for a direct test of the proposed factor structure rather than searching for the existence of a factor structure, as with exploratory factor analyses. With confirmatory factor analysis, testing theoretically meaningful alternative models lends credence to the conclusions one can draw from results (Lee & Hershberger, 1990; MacCallum et al., 1993), thus, we also compared the results of the hypothesized seven-factor model with one dedication factor, six constraint factors to a two-factor model that included one dedication factor and one global constraint factor.

Our data includes both members of each couple, which creates dependency in the data (i.e., observations are non-independent). This dependency violates basic assumptions for confirmatory factor analyses, so we followed guidelines provided by Kenny, Kashy, and Cook (2006) for dyadic confirmatory factor analyses. Following their approach, we correlated men's and women's error terms for each item to account for the systematic error (i.e., the expected influence of a partner on his/her partner's responses). The dyadic confirmatory factor analysis also includes item to latent factor loadings, correlations of error terms across men and women, correlations among latent factors for men and women separately, and latent factor correlations for each subscale within couples.

To test the first hypothesis, we first tested a model with seven factors and all 36 items. The error terms were correlated across men and women, but all paths were free to vary (i.e., were unconstrained) between men and women. Based on this model, we then excluded items that had factor loadings (e.g., multiple squared correlations) of lower than .30 for both men and women. For items that had above .30 for women or men but lower than .30 for the partner, we excluded the item if the partner's factor loading was below .10. Next, after deleting poor fitting items, we tested the factor invariance for men and women. To do so, we compared a model in which the paths were free to vary for men and women to a model where the paths were constrained to be equal. If the latter model demonstrates equal or better fit than the unconstrained model then it is reasonable to assume that the model fits similarly for men and women (Kenny et al., 2006).

In confirmatory factor analysis, adequacy of the proposed model as well as testing differences between models can be ascertained by examining fit statistics. Since there is little consensus about the most appropriate fit statistics for confirmatory factor analyses (Heubeck & Neill, 2000), we used a variety of indices. First, the Root Mean Square Error Approximation (RMSEA) was one measure of model fit used for the current study. RMSEA scores that are closer to 0 are preferred and estimates ranging from .05 to .08 are deemed adequate (Kenny & McCoach, 2003). Second, the Standardized Root Mean Residual (SRMR) is similar to the RMSEA in which lower values are preferred; SRMR of .10 or lower suggest an adequate fitting model (R. Kline, 1998). Third, the Comparative Fit Index (CFI) estimates should be.90 or higher for adequate model fit (R. Kline, 1998). A fourth fit statistic calculated here was chi-square divided by degrees of freedom (chi-square/df) with scores less than 3 indicating a “good” model fit (R. Kline, 1998). Lastly, we compared the chi-square estimates for the various models to determine which models were a better represented the data.

Results

Dyadic Confirmatory Factor Analysis

To test our first hypothesis that the factor structure would be a good representation of Stanley and Markman's (1992) theory of commitment (i.e., six constraint subscales measuring concern for partner welfare, social pressure, termination procedures, alternative available partners, alternative financial status, and structural investments and one subscale for dedication), we conducted a dyadic confirmatory factor analysis with both samples combined with all 36 items loading on seven factors. This model showed marginal fit (see Table 1). Based on our inclusion/exclusion criteria for items, we trimmed eleven items (one item from the Social Pressure subscale; one item from Concern for Partner Welfare; one item from Alternative Available Partners; two items from Structural Investments, and six items from Dedication). In all cases both men and women had squared multiple correlations lower than .30 for these items. The removal of these items reduced some of the subscales to two items, which may not be advisable in some situations (see Discussion). The seven-factor model with 25 items demonstrated acceptable fit on the majority of the fit indices. The unstandardized and standardized factor loadings for the 25 items in the unconstrained model are presented in Table 2. This seven-factor model demonstrated better fit than the two factor model with global constraint and global dedication only), χ2difference(50) = 1700.62, p < .001. The relative fit of the seven-factor model was 88% better than the null model; whereas the relative fit of the two-factor model was only 56% better than the null model.

Table 1.

Summary of Fit Indices for Dyadic Confirmatory Factor Analyses

Model χ 2 df χ2/df1 CFI SRMR RMSEA (90% CI)
7-factors, 36 items (Paths Unconstrained) 3785.73 2399 1.58 .80 .082 .043 (.040–.045)
7-factors, 25 items (Paths Unconstrained) 1769.37 1101 1.61 .88 .085 .044 (.040–.047)
7-factors, 25 items (Paths Constrained) 1796.12 1119 1.61 .87 .085 .044 (.040–.047)
2-factors, 25 items (Paths Constrained) 3496.74 1169 2.99 .56 .11 .079 (.076–.082)
Open in a new tab

Note. CFI = Comparative Fit Index, estimates of .90 or greater indicate good fit; SRMR = Standardized Root Mean, estimates of .10 or lower are preferred; RMSEA = Residual Root Mean Square Error Approximation, estimates of .05 or lower indicate good fit.

Table 2.

Items and Factor Loadings for Men and Women in the Trimmed Unconstrained Model

Women Men
Item Subscale b SE β b SE β
1. My friends would not mind if my partner and I broke up. Social Pressure Set to 1 .55 Set to 1 .58
2. If we ended this relationship, I would feel fine about my financial status. Financial Set to 1 .83 Set to 1 .63
3. The steps I would need to take to end this relationship would require a great deal of time and effort. Termination Set to 1 .74 Set to 1 .68
4. I could not bear the pain it would cause my partner to leave him/her even if I really wanted to. CPW Set to 1 .75 Set to 1 .65
5. It would be difficult for my friends to accept it if I ended the relationship with my partner. Social Pressure 1.17 .16 .52 1.10 .14 .56
6. It would be relatively easy to take the steps needed to end this relationship. Termination .77 .07 .70 .91 .09 .73
7. I would not have trouble supporting myself should this relationship end. Financial 1.01 .07 .82 1.13 .15 .75
8. My family really wants this relationship to work. Social Pressure 1.28 .15 .72 1.06 .12 .70
9. I would have trouble finding a suitable partner if this relationship ended. Availability Set to 1 .44 Set to 1 .53
10. I believe there are many people who would be happy with me as their spouse or partner. Availability .80 .13 .56 .81 .09 .64
11. I have put a number of tangible, valuable resources into this relationship. Investments Set to 1 .66 Set to 1 .75
12. Though it might take awhile, I could find another desirable partner if I wanted or needed to. Availability 1.40 .24 .85 1.46 .18 .99
13. I would not have any problem with meeting my basic financial needs for food, shelter, and clothing without my partner. Financial .85 .06 .81 .74 .10 .58
14. I have put very little money into this relationship. Investments 1.06 .29 .73 .75 .21 .61
15. The process of ending this relationship would require many difficult steps. Termination .90 .08 .76 1.08 .11 .79
16. If I really felt I had to leave this relationship, I would not be slowed down by concerns for how well my partner would do without me. CPW .75 .11 .61 .68 .13 .46
17. My family would not care if I ended this relationship. Social Pressure 1.58 .18 .83 .99 .11 .73
18. My relationship with my partner is more important to me than almost anything in my life. Dedication Set to 1 .72 Set to 1 .45
19. I want this relationship to stay strong no matter what rough times we encounter. Dedication .77 .06 .82 .64 .09 .73
20. I like to think of my partner and me more in terms of “us” and “we” than “me” and “him/her.” Dedication .76 .08 .56 .81 .12 .61
21. I think a lot about what it would be like to be married to (or dating) someone other than my partner. Dedication .94 .10 .57 .99 .14 .63
22. My relationship with my partner is clearly part of my future life plans. Dedication .73 .06 .76 .77 .10 .74
23. My career (or job, studies, homemaking, childrearing, etc.) is more important to me than my relationship with my partner. Dedication .89 .08 .67 .74 .12 .53
24. I do not want to have a strong identity as a couple with my partner. Dedication .69 .07 .55 .73 .12 .52
25. I may not want to be with my partner a few years from now. Dedication 1.01 .08 .76 1.01 .14 .74
Open in a new tab

Notes. All factor loadings were significant at p < .001. CPW = Concern for Partner's Welfare; Financial = Financial Alternatives; Termination = Termination Procedures; Investments = Structural Investments; Availability = Availability of Other Partners.

Next, we tested our second hypothesis that the seven-factor structure of the Commitment Inventory would fit for men and women, through a two-step process. We first tested a model in which the paths were constrained to be equal across men and women. We then compared this model to a model in which paths were free to vary across men and women. The constrained model demonstrated better fit than the unconstrained model, χ2difference(18) = 26.75, p < .05. This suggests that the factor structure was similar for men and women, supporting our second hypothesis.2

Commitment Inventory Subscale Correlations

The relationships between the subscales of the Commitment Inventory and relationship adjustment, negative communication, and religiosity were examined as indicators of criterion-related validity for the third hypothesis (Table 3). The correlations between the Revised Commitment Inventory subscales and the relationship adjustment measures (i.e., MAT and DAS) were small to medium; as predicted the strongest of these relationships were between relationship adjustment and Dedication. The small correlations between relationship adjustment and constraint commitment subscales illustrate divergent validity.

Table 3.

Correlations among Revised Commitment Inventory Subscales and Other Measures

1 2 3 4 5 6 7 8 9 10 11
1 Dedication -- .12 −.07 .12 .50** .13 .22** .57** .42** −.40** .33**
2 CPW .15* -- .06 .42** .28** .09 .15* −.06 .08 −.03 .06
3 Financial .07 .19** -- −.02 −.02 −.07 .24** −.09 −.06 .05 −.07
4 Termination .17* .42** .09 -- .17* .10 .05 −.02 .12 −.03 .06
5 Social Pressure .45** .30** .10 .29** -- .03 .22** .16 .30** −.28** .27**
6 Investments .07 .08 −.03 .18** .02 -- −.09 .08 −.04 .01 .06
7 Availability .22** .31** .23** .21** .25** −.06 -- .17 .29** −.12 .04
8 DAS .23** −.14 −.19 −.10 .34** .09 .12 -- -- −.67** .13
9 MAT .54** .11 −.03 −.01 .27** −.07 .13 -- -- −.63** .03
10 Neg. Comm. −.38** .01 .08 .10 −.31** .04 −.13 −.64** −.59** -- −.14*
11 Religiosity .29** .12 −.01 .05 .11 .02 .16* .02 .05 −.10 --
Open in a new tab

Notes. Correlations for women are below the diagonal, correlations for men are above the diagonal; CPW = Concern for Partner Welfare; Financial = Financial Alternatives; Termination = Termination Procedures; Investments = Structural Investments; Availability = Availability of Other Partners; DAS = Dyadic Adjustment Scale; MAT =Martial Adjustment Test; Neg. Comm. = Negative Communication.

*

p < .01,

**

p < .001.

Negative communication also correlated most strongly with Dedication (r = −.38 for women and r = −.40 for men). Further, negative communication was negatively correlated with Social Pressure (r = −.31 for women and r = −.28 for men). These results suggest that as couples communicate more negatively, they report less dedication to the relationship and perceive less pressure from family and friends to stay in the relationship. As predicted, religiosity was significantly related to Dedication, but demonstrated null correlations with the other Revised Commitment Inventory subscales, with two exceptions. Women's religiosity was correlated with Available Alternative Partners (r = .16) and men's religiosity was correlated with Social Pressure (r = .27).

The bivariate correlations among the Revised Commitment Inventory subscales were small to medium (see Table 3). This finding suggested that the subscales assess different, yet related aspects of commitment. None of the Commitment Inventory subscale correlations were large enough to indicate that the subscales were measuring the same construct.

Within Couple Correlations

Examining the within-couple correlations from the confirmatory factor analysis generally revealed small to medium relationships (see Table 4). The largest within-couple correlation was for Dedication; the smallest was for Alternative Financial Status, which was non-significant.

Table 4.

Within Couple Correlations for the Revised Commitment Inventory Subscales

Women
1 2 3 4 5 6 7
Men 1. Dedication .45** .06 .03 .13 .27** −.05 .22**
2. CPW .20** .21** .09 .11 .15* −.01 .12
3. Financial −.10 .06 .06 .02 −.06 .08 .19**
4. Termination .16* .08 .07 .14* .11 .08 .01
5. Social Pressure .30** .14* −.03 .13 .29** .01 .10
6. Investments .06 .07 .11 .13 .05 .14* .04
7. Availability .16* .13 .05 .13 .13 .03 .15*
Open in a new tab

Notes. These correlations account for the dependency within couples (i.e., error terms correlated). CPW = Concern for Partner Welfare; Financial = Financial Alternatives; Termination = Termination Procedures; Investments = Structural Investments; Availability = Availability of Other Partners.

*

p < .01

**

p < .001

Internal Consistency

Estimates of internal consistency (i.e., Cronbach's alpha) are determined by the inter-item correlations and the number of items including in the scale (Cronbach, 1951). Accordingly, there is a positive relationship between the number of items on a scale and the Cronbach's alpha coefficient (Cortina, 1993). Despite the low number of items for the majority of the scales, we found that nine of the fourteen had alpha coefficients greater than .70, a common but rather arbitrary cut-point (Gable & Wolfe, 1993). Available Alternative Partners had a low alpha for women (.59) but not for men (.72). Concern for Partner Welfare only had two items and the alpha was low women (.63) and especially low for men (.46). Similarly, Structural Investments demonstrated alphas in the low .60s for men and women. Researchers should be cautious in their use of these of these specific subscales (see below).

Discussion

This study explored the factor structure and internal consistency of a briefer, updated version of Stanley and Markman's (1992) Commitment Inventory in samples of engaged and cohabiting couples. A dyadic confirmatory factor analysis allowed for new tests of the measure's structure using data from couples. In general, findings support confidence in the use of the Revised Commitment Inventory. After trimming poor-fitting items, the Revised Commitment Inventory consists of 25 items loading on seven factors (six constraint subscales and one dedication subscale). Correlations with other measures support the validity of this inventory, however researchers should be cautioned that some subscales only contain two items, which compromised their internal consistency. Future research could add additional items to these subscales to increase internal consistency.

The results of the dyadic confirmatory factor analyses suggested that the seven-factor structure of the Commitment Inventory was a good fit for the data and better than a two-factor model in which all constraint items were placed on a single global constraint factor. This finding supports the use of the specific constraint subscales over the total constraint commitment score. At the same time, the total constraint scale demonstrated reasonable internal consistency, so some researchers may find it useful, depending on the research questions. Oftentimes it will be important to examine specific areas of constraint and relative contributions of different areas for relationships outcomes. For example, future research could test whether financial pressures or concern for a partner's welfare tend to be more potent predictors of remaining in relationship and whether there are individual differences in these associations.

The use of a shortened (8-item) dedication subscale was supported, which is consistent with prior recommendations that dedication can be assessed globally (Stanley & Markman, 1992). The items that were removed from the global dedication subscale were primarily related to meta-commitment and sacrifice. Potentially, these items were less salient for premarital and cohabiting couples than the others.

Internal consistency tended to bein the marginal to acceptable range for most subscales. These results are typical for scales that have a limited number of items (Cortina, 1993; Helms et al., 2006). Concern for Partner Welfare and Structural Investments, both subscales with only two items, demonstrated the lowest internal consistency. Additionally, Available Alternative Partners had low internal consistency for women but not for men. Generally, reliability is a lower-bound estimate of reliability and low reliability affects Type II error (not Type I), suggesting that low internal consistency will make it less likely that significant results will be observed (if they truly exist), but this type of error will not result in spuriously significant findings (Badr & Acitelli, 2008; Rosenthal, 1995). In fact, reliabilities as low as .50 have been shown to not seriously affect validity (Schmmitt, 1996). Thus, the correlations found with other relationship measures likely demonstrate true effects. Future work should consider adding items to the scales that have low reliability to increase Cronbach's alphas.

The factor structure was similar between men and women, suggesting that the ways in which men and women interpreted the items were comparable, as least in relation to other items on the measure. This finding is important to research on gender differences and commitment (e.g., Rhoades et al., 2006; Stanley et al., 2004). When gender differences in mean levels of commitment are found, they can be interpreted with more confidence, because an alternative hypothesis that the gender differences are due to differences in how men and women define and subsequently score on a measure of commitment can be ruled out. However, our study did not address if men and women have different conceptions in the fundamental definition of commitment. That is, we did not examine if men and women agreed that the items within the Commitment Inventory are consistent with they way they define commitment. To our knowledge, there is no research that has examined this issue.

Additionally, the current study found that partners' scores on the same subscales of the Commitment Inventory were significantly correlated. The only subscale on which partners' scores were not significantly correlated was Financial Alternatives. Knowing how an individual views his or her alternative financial situation appears to be fairly unrelated to how the partner might view his or her own potential financial situation should the relationship end. This non-significant finding was not surprising given that partners could very reasonably have different financial prospects should they break up. For example, in many couples, one partner (often the male) earns more income than the other. If such a relationship were to end, the one who makes more money or who has greater potential to generate income would be in a far better position to continue to support him or herself.

As expected, associations among subscales of the Commitment Inventory and relationship adjustment and negative communication showed basic evidence of criterion-related validity. Specifically, relationship adjustment and negative communication were associated with dedication in expected directions and to a lesser extent with social pressure, which is consistent with prior research (Adams & Jones, 1997; Johnson et al., 1999; Stanley & Markman, 1992). One interpretation of these data is that as couples' negative communication increases, their intrinsic desire to be with their partner decreases and the pressure they perceive from others to stay in the relationship also decreases. While these data are correlational and causality cannot be determined, there is likely a bi-directional effect between these forms of commitment and negative communication (as well as relationship adjustment). Most of the constraint subscales exhibited small to null correlations with relationship adjustment and negative communication. This finding is consistent with commitment theory, insofar as constraints are not necessarily negative or positive (Stanley & Markman, 1992).

Findings regarding religiosity were also in line with previous research. For both men and women, religiosity was significantly associated with dedication, as was found in the initial study of the Commitment Inventory (see Stanley & Markman, 1992). For men, but not for women, there was also a significant association between religiosity and perceived social pressure. There was a significant association between social pressure and religiosity in the original study on the Commitment Inventory as well, but it was not broken down by gender, nor was it tested in an unmarried sample. We had not predicted an association between social pressure and religiosity because we assumed that social pressure would be relevant only to those who were already married because of values about marriage and divorce among friends and family. In actuality, religiosity does seem relevant to perceived social pressure, at least for men. Future research could examine the meaning of social pressure in married versus unmarried samples more directly and could also help to better understand why there may be a related gender difference.

Limitations and Future Directions

There are limitations to this study that could be addressed with future studies. First, the purpose of our study was to examine commitment in unmarried couples, thus, the generalizability of our results to couples in other stages in a relationship (e.g., early dating or married) is limited. Further, there were differences in the mean scores between the two samples, which is to be expected given the differences in relationship stage across the two samples. There is a benefit to this heterogeneity in that increases the external validity of the study, however, questions remain about whether the factor structure would replicate in either one of these samples alone. Larger samples will be needed to generate enough statistical power for these tests. Second, our sampling method was based convenience and subsequently may not be nationally representative. Third, our intent was not to validate the measure through a full construct and criterion-related validity analysis. Rather, we wanted to focus primarily on the factor structure of the measure, which was supported. Stanley and Markman (1992) demonstrated adequate criterion-related validity for the majority of the subscales, however, future research should continue to address the validity of the Revised Commitment Inventory with other samples and longitudinally. Lastly, based on model fit, the use of the global constraint scale was not supported in the current study; however, future research should examine if a measure that sums across various sources of constraints could be useful. For example, a checklist (yes/no) of current constraints maybe conceptually meaningful as some couples will likely have more constraints, but the relative importance of having that constraint may not be linear. That is, the step from not having a constraint (e.g., not concerned about financial welfare of partner) to having that constraint may be a meaningful step.

Table 5.

Subscale Means, Standard Deviations, and Internal Consistency Estimates for Men and Women

Mean (SD) Internal Consistency
Subscale Women Men Women Men
Dedication 6.30(0.81) 6.03(0.86) .86 .81
Concern for Partner 4.46(1.54) 4.84(1.28) .63 .46
Financial Alternatives 2.83(1.59) 2.39(1.29) .86 .69
Termination Procedures 5.38(1.38) 5.41(1.31) .77 .77
Available Altrn Partners 2.96(1.21) 3.09(1.30) .59 .72
Social Pressure 5.41(1.14) 5.32(1.08) .74 .73
Structural Investments 4.86(1.36) 5.17(1.24) .64 .63
Total Constraint 4.32(0.77) 4.37(0.63) .77 .69
Open in a new tab

Acknowledgments

Funding The author(s) disclosed receipt of the following financial support for the research and/ or authorship of this article: Preparation of this manuscript was supported in part by grants from The National Institute of Child Health and Human Development to the second, third, and fourth authors (5R01HD047564 and R01HD053314). The contents are solely the responsibility of the authors and do not necessarily represent the official views of NIH or NICHD. The contents are solely the responsibility of the authors and do not necessarily represent the official views of National Institutes of Health or National Institute of Child Health and Human Development.

Footnotes

Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the authorship and/or publication of this article.

1

T-tests comparing those who were included in the engaged sample for the current study (n = 200 couples) to couples who were not included (either because they did not complete post or because of missing data on the Commitment Inventory; n = 42 couples) revealed that men and women who were included were significantly older and more educated than men and women who were not included, ps < .05. However, there were no significant differences between the groups in terms of income level, Marital Adjustment Test scores, or Dedication subscale scores (measured at the first assessment). The constraint subscales were not assessed during the first assessment.

2

The cohabiting sample had significantly higher alphas on some subscales than the engaged sample. The largest difference for women was on the dedication subscale (engaged sample alpha: .72; cohabiting sample: .86, p(difference) < .001). The largest difference for men was on the social pressure subscale (engaged sample alpha: .63, cohabiting sample .77, p(difference) = .01). We attempted a multiple groups dyadic confirmatory factor analysis to determine if there were differences in the model fit across the two samples; however, the model did not converge. We also tested for other differences across the two samples. Using a Bonferroni corrected p-value (p = .006), we found that the engaged sample had higher religiosity, dedication, and social pressure than the cohabiting sample. We also found that concern for partner welfare was higher in the engaged sample (only for women) than in the cohabiting sample and that alternative availability was higher in the cohabiting sample than in the engaged sample.

References

  1. Adams JM, Jones WH. The conceptualization of marital commitment: An integrative analysis. Journal of Personality and Social Psychology. 1997;72(5):1177–1196. [Google Scholar]
  2. Badr H, Acitelli LK. Attachment insecurity and perceptions of housework: Associations with marital well-being. Journal of Family Psychology. 2008;22:313–319. doi: 10.1037/0893-3200.22.2.313. [DOI] [PubMed] [Google Scholar]
  3. Bramlett MD, Mosher WD. Cohabitation, marriage, divorce, and remarriage in the United States (No. 23) National Center for Health Statistics; Hyattsville, MD: 2002. [PubMed] [Google Scholar]
  4. Cortina JM. What is coefficient alpha? An examination of theory and applications. Journal of Consulting and Clinical Psychology. 1993;56:754–761. [Google Scholar]
  5. Crane DR, Allgood SM, Larson JH, Griffin W. Assessing marital quality with distressed and nondistressed couples: A comparison and equivalency table for three frequently used measures. Journal of Marriage and Family. 1990;52(1):87–93. [Google Scholar]
  6. Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika. 1951;16:297–334. [Google Scholar]
  7. Demoss Y. Brief interventions and resiliency in couples. Dissertation Abstracts International: Science B: The Sciences and Engineering. 2004;65(5-B):2619. [Google Scholar]
  8. Ferguson J. Factors contributing to satisfying long-term marriages. Dissertation Abstracts International. 1993;54(5-B):2808. [Google Scholar]
  9. Gable RK, Wolfe MB. Instrument development in the affective domain: Measuring attitudes and values in corporate and school settings. 2nd Ed Kluwer Academic Publishers; Norwell, MA: 1993. [Google Scholar]
  10. Heubeck B, G., Neill JT. Confirmatory factor analysis and reliability of the mental health inventory for Australian adolescents. Psychological Reports. 2000;87(2):431–440. doi: 10.2466/pr0.2000.87.2.431. [DOI] [PubMed] [Google Scholar]
  11. Helms J, Henze KT, Sass TL, Mifsud VA. Treating Cronbach's alpha reliability coefficients as data in counseling research. The Counseling Psychologist. 2006;34(5):630–660. [Google Scholar]
  12. Hunsley J, Best M, Lefebvre M, Vito D. The seven-item short form of the Dyadic Adjustment Scale: Further evidence for construct validity. American Journal of Family Therapy. 2001;29(4):325–335. [Google Scholar]
  13. Impett EA, Beals KP, Peplau LA. Testing the investment model of relationship commitment and stability in a longitudinal study of married couples. Current Psychology. 2002;20(4):312–326. [Google Scholar]
  14. Johnson MP. Commitment: A conceptual structure and empirical application. Sociological Quarterly. 1973;14:395–406. [Google Scholar]
  15. Johnson MP, Caughlin JP, Huston TL. Tripartite nature of marital commitment: Personal, moral, and structural reasons to stay married. Journal of Marriage and Family. 1999;61:160–177. [Google Scholar]
  16. Kline GH, Stanley SM, Markman HJ, Olmos-Gallo PA, St. Peters M, Whitton SW, et al. Timing Is everything: Pre-engagement cohabitation and increased risk for poor marital outcomes. Journal of Family Psychology. 2004;18(2):311–318. doi: 10.1037/0893-3200.18.2.311. [DOI] [PubMed] [Google Scholar]
  17. Kline RB. Principles and practice of structural equation modeling. The Guilford Press; New York: 1998. [Google Scholar]
  18. Kenny DA, Kashy DA, Cook WL. Dyadic data analysis. Guilford Press; New York NY: 2006. [Google Scholar]
  19. Kenny DA, McCoach DB. Effect of the number of variables on measures of fit in structural equation modeling. Structural Equation Modeling. 2003;10(3):333–351. [Google Scholar]
  20. Lee S, Hershberger S. A simple rule for generating equivalent models in structural equation modeling. Multivariate Behavioral Research. 1990;25:313–334. doi: 10.1207/s15327906mbr2503_4. [DOI] [PubMed] [Google Scholar]
  21. Locke HJ, Wallace KM. Marital-adjustment and prediction tests: Their reliability and validity. Marriage and Family Living. 1959;21:251–255. [Google Scholar]
  22. MacCallum RC, Wegener DT, Uchino BN, Fabrigar LR. The problem of equivalent models in applications of covariance structure analysis. Psychological Bulletin. 1993;114:185–199. doi: 10.1037/0033-2909.114.1.185. [DOI] [PubMed] [Google Scholar]
  23. Manning WD, Smock PJ. Measuring and modeling cohabitation: New perspectives from qualitative data. Journal of Marriage and Family. 2005;67:989–1002. [Google Scholar]
  24. Markman HJ, Whitton SW, Kline GH, Thompson H, St. Peters M, Stanley SM, et al. Use of an empirically-based marriage education program by religious organizations: Results of a dissemination trial. Family Relations. 2004;53:504–512. [Google Scholar]
  25. Owen J, Rhoades G, Stanley SM, Fincham FD. “Hooking up” among college students: Demographic and psychosocial correlates. Archives of Sexual Behavior. doi: 10.1007/s10508-008-9414-1. in press. [DOI] [PubMed] [Google Scholar]
  26. Pistole C, M., Vocaturo LC. Attachment and commitment in college students' romantic relationships. Journal of College Student Development. 1999;40(6):710–720. [Google Scholar]
  27. Rhoades GK, Stanley SM, Markman HJ. Couples' reasons for cohabitation: Associations with individual well-being and relationship quality. Journal of Family Issues. 2009;30:233–258. doi: 10.1177/0192513X08324388. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Rhoades GK, Stanley SM, Markman HJ. Pre-engagement cohabitation and gender asymmetry in marital commitment. Journal of Family Psychology. 2006;20:553–560. doi: 10.1037/0893-3200.20.4.553. [DOI] [PubMed] [Google Scholar]
  29. Rosenthal R. Methodology. In: Tesser A, editor. Advanced social psychology. McGraw-Hill; Boston: 1995. pp. 17–49. [Google Scholar]
  30. Schmitt N. Uses and abuses of coefficient alpha. Psychological Assessment. 1996;8:350–353. [Google Scholar]
  31. Spanier GB. Measuring dyadic adjustment: New scales for assessing the quality of marriage and similar dyads. Journal of Marriage and the Family. 1976;38:15–28. [Google Scholar]
  32. Spiers CJ. Commitment and stability in lesbian relationships. Dissertation Abstracts International: Section B: The Sciences and Engineering. 1998;59(6-B):3076. [Google Scholar]
  33. Stanley SM. What is it with men and commitment, anyway?. Keynote address to the 6th Annual Smart Marriages Conference; Washington D.C.. 2002. [Google Scholar]
  34. Stanley SM, Amato PR, Johnson CA, Markman HJ. Premarital education, marital quality, and marital stability: Findings from a large, random, household survey. Journal of Family Psychology. 2006;20:117–126. doi: 10.1037/0893-3200.20.1.117. [DOI] [PubMed] [Google Scholar]
  35. Stanley SM, Markman HJ. Assessing commitment in personal relationships. Journal of Marriage and Family. 1992;54(3):595–608. [Google Scholar]
  36. Stanley SM, Rhoades GK, Markman HJ. Sliding vs. deciding: Inertia and the premarital cohabitation effect. Family Relations. 2006;55:499–509. [Google Scholar]
  37. Stanley SM, Markman HJ. Marriage in the 90s: A nationwide random phone survey. PREP; Denver, CO: 1997. [Google Scholar]
  38. Stanley SM, Markman HJ, Whitton SW. Communication, conflict, and commitment: Insights on the foundations of relationship success from a national survey. Family Process. 2002;41:659–675. doi: 10.1111/j.1545-5300.2002.00659.x. [DOI] [PubMed] [Google Scholar]
  39. Stanley SM, Markman HJ, Prado LM, Olmos-Gallo PA, Tonelli L, St. Peters M, et al. Community-based premarital prevention: Clergy and lay leaders on the front lines. Family Relations. 2001;50(1):67–76. [Google Scholar]
  40. Stanley SM, Whitton SW, Markman HJ. Maybe I do: Interpersonal commitment and premarital or nonmarital cohabitation. Journal of Family Issues. 2004;25(4):496–519. [Google Scholar]
  41. Stanley SM, Whitton SW, Sadberry SL, Clements ML, Markman HJ. Sacrifice as a predictor of martial outcomes. Family Process. 2006;45:289–303. doi: 10.1111/j.1545-5300.2006.00171.x. [DOI] [PubMed] [Google Scholar]
  42. Whitton SW, Stanley SM, Markman HJ. If I help my partner, will it hurt me? Perceptions of sacrifice in romantic relationships. Journal of Social and Clinical Psychology. 2006;26:64–91. [Google Scholar]

RESOURCES