Development and Validation of the Sorokin Psychosocial Love Inventory for Divorced Individuals

Abstract

Objective:

This study describes the development and validation of the Sorokin Psychosocial Love Inventory (SPSLI) measuring love actions toward a former spouse.

Method:

Classical measurement theory and confirmatory factor analysis (CFA) were utilized with an a priori theory and factor model to validate the SPSLI.

Results:

A 15-item scale containing five subscales (intensity, extensity, purity, adequacy, and duration) showed good fit, statistically significant factor loadings, good reliability, and validity. Multigroup CFA indicated both measurement equivalence and structural equivalence between males and females.

Discussion:

The SPSLI supports Sorokin’s multidimensional theory of love and provides clinicians a method to measure love actions of divorcing individuals. Implications for practice and future research are discussed.

Keywords

love divorce scale development scale validation exploratory and confirmatory factor analysis

Dissolution of marriage, more commonly called divorce, is here to stay even though a concerted effort has been made to curb its effect on society ((Brotherson & Duncan, 2004; Huston & Melz, 2004). Although it is difficult to determine a complete count of divorces in the United States, due to a lack of consistent state reporting, it is safe to approximate that 50% of marriages will be voluntarily disrupted by divorce or separation (Amato, 2010). Of those dissolutions, half involve a child under the age of 18 years (Amato, 2000). When divorce occurs, it is typically viewed as a fatal relationship failure that dramatically impacts not only the couple but also their children and extended families. It also creates repercussions in the community by endangering institutional and economic stability (Wilcox, Marquardt, Popenoe, & Whitehead, 2010).

The divorce phenomenon is a complex and multidimensional process often marked by instability (Potter, 2010), including a 3.7-fold increased risk in mood disorders, a 2.5-fold increased risk in anxiety disorders, and a 3.3-fold increased risk in substance use disorders (Chatav & Whisman, 2007). Compared to married individuals, those people experiencing divorce report more symptoms of anxiety, depression, substance abuse, and a greater overall mortality (Bierman, Fazio, & Milkie, 2006; Hughes & Waite, 2009; Waite, Luo, & Lewin, 2009; Williams & Umberson, 2004; Zhang & Hayward, 2006). Divorce produces emotional crisis in individuals that often leads to conflict, which occurs before, during, and after the divorce process (Booth & Amato, 2001). Divorce also causes posttraumatic stress for many which lead to postdivorce conflict (Chung et al., 2003). Over the past decade, research continues to show that children whose parents divorced scored lower on average on a variety of emotional, behavioral, social, health, and academic outcomes (Frisco, Muller, & Frank, 2007; Hango & Houseknecht, 2005; Sun & Li, 2002; Troxel & Matthews, 2004; Wallerstein, 2005). Family conflict was specifically found to have more negative effects on the well-being of children than divorce or separation (Amato, 2010; Kelly & Emery, 2003).

The majority of marital separations end rather quickly in reconciliation or dissolution (Amato, 2010). For others, the path is met with conflict and turmoil (Levite & Cohen, 2012). However, even for those couples that choose a more peaceful path, exogenous influences can lead them to look at their relationship dissolution as a failure (Duck, 2007). This does not leave much room to divorce and still show love actions toward a former spouse. Looking at divorce from a strengths perspective, as has been promoted in social work (Saleeby, 2009) and positive psychology (Peterson, 2006), may be able to correct the imbalance and challenge the assumptions of the disease model that divorce seems to follow. This does not preclude acknowledgment of the anger, conflict, and sadness that occur during the divorce process but instead focuses on a salutogenic process that looks at the human capacity to love even in the midst of what is for some the most devastating time in their lives. The adoption of a strengths perspective by assessing for love actions for divorced individuals may be the impetus that allows marital relationships to end with understanding, generosity, humility, and other-regarding love actions. A strengths approach promotes personal and community capacity and counters the focus on problems and pathology (Fawcett & Reynolds, 2010). It is not blindly optimistic or illusory to believe that people can treat each other with other-regarding love during difficult times. In order to address love actions during divorce, it is necessary to have a theory that operationalizes love so that it can be measured.

Sorokin’s Love Theory

Although there are many taxonomies of love, romantic love has generally been promoted in society and science over the past decades (Fehr, 2006; Hatfield & Sprecher, 1986; Hendrick & Hendrick, 1986; Lasswell & Lasswell, 1976; Levin, 2000; Masuda, 2003; Rubin, 1970). While much has been written about love, marriage, divorce, and close relationships (Berscheid, 2010), including attachment (Hazan & Shaver, 1987; Johnson, 2004; Saini, 2012), forgiveness (Bonarch & Sales, 2002; Rohde-Brown & Rudestam, 2011), and more complex constructs such as hope, virtue, character (Linley, Joseph, Harrington, & Wood, 2006), empathy (de Waal, 2009), sympathy (Gerdes, Lietz, & Segal, 2011), altruism (Post, 2003; Sorokin, 1950), and compassion (Underwood, 2009), an understanding of love has continued to confound scholars although it has surfaced in discussions related to the above concepts. In order to utilize an understanding of love that is multidimensional, this scale development and validation study is guided by a theory of love developed by Pitirim Sorokin in the 1950s (Sorokin, 1954b). It includes an operationalized five-factor model that can be used as an experimental tool to research core questions about other-regarding love. While there is no one model that has evolved as a panacea, this theory and model provide the basis for development of a scale that explores a way to measure love guiding humans to a more compassionate way of dealing with each other, especially during divorce (Sorokin, 1954a, 1954b). Conceptualizing love through Sorokin’s lens allows the focus to remain on “love,” a strengths construct that is intrinsic in human beings. This prevents the focus to be diverted to either emotions (compassion, sympathy, and empathy) or altruism, which are easier to conceptualize. Both the social work strengths perspective and positive psychology support the idea that human goodness not only exists but is achievable if addressed (Peterson, 2006; Saleeby, 2009). What makes this theory applicable to us today is the ease with which it is understood and applied in close personal relationships. Also, we, like other scholars (Levin & Kaplan, 2010; Post, 2003) wanted to build on Sorokin’s work that he started many decades ago and determine whether his original love model could be developed into a scale that can be validated and used to promote love within difficult life situations.

After a critical review of Sorokin’s publications and other Sorokin scholars, we developed the following operational definitions for each of the five love constructs that form part of Sorokin’s theory (Levin & Kaplan, 2010; Post, 2003; Sorokin, 1950, 1954a, 1954b, 1958, 1963, 1965): Intensive love is other-regarding actions that range from little loss to self to great loss to self. Extensive love is other-regarding actions starting with the love of oneself, extending to family and friends and extending further toward all human beings, without regard for who they are and how different their actions are from ours. Pure love refers to other-regarding actions that range from impure love that is but a means to a selfish end to other-regarding actions that are motivated by love alone without expectations. Adequate love is other-regarding actions ranging from actions where the subjective motive is loving, but the objective consequence is nonloving or the subjective motive is nonloving, but the objective consequence is loving to wise and creative other-regarding actions that are both subjectively and objectively loving and in unity. Duration love is other-regarding actions that span from the shortest possible moment to the whole life of an individual.

Item Generation

The item pool for the different domains was designed to follow the principles of the domain-sampling model of measurement, in which a theoretical infinite number of items may be used to create a scale that measures the different constructs. The skill lies in choosing a sample that best represents the population (Nunnally & Bernstein, 1994). Four items per construct were developed, the upper bound of the preferred minimum amount of items needed to confirm potentially stand-alone, unifactor subscales (Hair, Black, Babin, & Anderson, 2010), for a total of 20 items in the multidimensional Sorokin Psychosocial Love Inventory (SPSLI). The pool of items was then sent to an expert panel consisting of 16 members, namely, 2 Sorokin experts, 4 academics, 3 people with master’s degrees, 3 undergraduates, and 4 divorcees. Typically only five experts are recommended to review a proposed instrument to detect marginal or bad items (Netemeyer, Bearden, & Sharma, 2003), but because of the complexity of the theory more were recruited. Reviewers rated each item in terms of low, moderate, or high relevance to the construct as well as clarity. Responses were compiled, and items with low ratings were reevaluated and rewritten. Reviewers also provided qualitative feedback. Based on this input, focus was placed on ensuring a better distinction between the five constructs. The item pool was also rewritten to show “actions” rather than “feelings” to align better with the intent of Sorokin’s theory. Further, items with the focus on “desires” of a former spouse were changed to “needs” to allow for more realistic items in terms of what can be expected within a divorce situation. A total of 18 of the 20 items were reworded as a result of the expert review. After the items were developed, they were scaled on a 5-point Likert-type scale ranging from strongly disagree (1) to strongly agree (5).

The purpose of this study was to perform a psychometric analysis of the aforementioned instrument to measure love actions toward a former spouse during or after a divorce. Specifically, it evaluated the factor structure of the proposed scale and tested whether the scale is representative of the underlying theory as developed by Sorokin.

Method

Design

This study was a cross-sectional survey design. The design was cross sectional because the data were gathered at essentially one point in time and was contextual because the individuals participating had all been divorced or were currently going through a divorced. The University Institutional Review Board approved the study.

Sampling and Data Collection Procedures

The survey was administered to volunteers who attended a court-mandated “Families in Transition Program” (FIT) while going through a divorce. Leaders of the FIT training program administered the survey at the beginning of their psychoeducational program that was offered at four locations over the study period. Additional participants were recruited via a secure online survey using purposeful sampling of individuals who either were going through a divorce or had been divorced. The online survey was advertised in a University’s daily e-mail notice to employees, faculty, and students two times over the course of 3 months, sent to coworkers at a social service delivery agency, sent to church members of various churches, advertised on social media including Facebook and Twitter, and also e-mailed to national divorce support groups, as well as a men’s group. The FIT program provided 149 (28%) participants and 381 (72%) individuals participated in the online survey for a total of 530 respondents.

The survey consisted of a consent preamble, a demographic form, the newly developed SPSLI with its 20 items, and 12 other additional instruments on attorney influence, equity, attachment styles, predisposition toward anger, anger toward spouse, altruism, empathy, compassion, sympathy, collectivism, and spirituality that was used as part of larger study to develop a conceptual model explaining the ability to show love actions toward a former spouse.

Analysis

Reliability and Content Validity on the Item Level for Subscales as Unidimensional Constructs

For the first part of the analysis, we confirmed the unidimensional latent constructs for each of the love domains, using traditional psychometric analysis, specifically exploratory factor analysis (EFA) with maximum-likelihood factor extraction using the software package IBM SPSS 20.0.0 (SPSS, 2011). We allowed these latent constructs to be reduced to no less than 3-item constructs based on the recommendations from Hair, Black, Babin, and Anderson (2010). We evaluated the internal consistency reliability of the different subscales (the overall proportion of true score variance to total observed variance) with the Cronbach’s α, using a criterion of .80 to indicate internal consistency. The validity of the unidimensional constructs were evaluated by examining the inter-item correlation (IIC) matrix and making sure they were ≥.30, examining the corrected item total correlations (ITC) to assure they were ≥.45 and determining the mean of all corrected ITC to assure that they were ≥.50 (DeVellis, 2012; Faul & van Zyl, 2004). Face validity was assumed after the completion of the expert panel review.

Testing the Underlying Factor Structure With an EFA

After we confirmed the unidimensional latent constructs for each domain of the SPSLI, the second part of the analysis focused on empirically appraising the underlying factor structure with an EFA, using maximum likelihood factor extraction and a promax oblique rotation (based on the data being normally distributed and the theoretical assumption that factors would be intercorrelated) (Costello & Osborne, 2005). Scale factorability was assessed with the Kaiser-Meyer-Olkin (KMO) measure of sample adequacy, using a criterion of 0.60 to indicate the existence of factors in a correlation matrix. The appropriateness of the oblique rotation was assessed with the factor correlation matrix, using a criterion of ≥0.32, indicating more than 10% overlap in variance among factors. Item deletion decisions were based on item communalities after rotation >0.40, absolute values of item loadings >0.45 and cross loadings with values >0.32 on at least two factors (Tabachnick & Fidell, 2012; Worthington & Whittaker, 2006).

Confirming the EFA Solution With a First-Order Confirmatory Factor Analysis (CFA)

The focus of this part of the analysis was to confirm the factor structure reliability and validity of the SPSLI, as determined by the EFA solution. The CFA analysis involved simultaneously estimating in oblique form (assuming all factors were correlated) all five measurement models (factors) through a freed symmetric matrix of correlated indicator error variances, using the software package IBM AMOS 20.0.0 (IBM Corporation, 2011). In the first-order CFA, we made no specific claim about the directions or patterns of factor interrelationships. We assessed model fit with the following goodness-of-fit indices: overall model fit with the relative likelihood ratio (χ²/df), the absolute goodness-of-fit index (GFI) that compares the model against no model, the comparative fit index (CFI) that compares the model against an independence model, the root mean square error of approximation (RMSEA) that investigates how well the model fits the population covariance matrix, and the standardized root mean square residual (SRMR) that represents the average discrepancy between the sample observed and hypothesized correlation matrices (Byrne, 2010). The recommended goodness-of-fit criteria used to evaluate these indices were χ²/df < 3 (Wheaton, Muthen, Alwin, & Summers, 1977); GFI and CFI ≥ 0.92 (revised criteria for multifactor instruments; Marsh, Hau, & Wen, 2004); RMSEA < 0.06 (Hu & Bentler, 1999); and SRMR < 0.08 (Hu & Bentler, 1995). We investigated model misspecification by evaluating the standardized residuals and the modification indices (Byrne, 2010).

We evaluated the reliability of the different subscales with a composite reliability (CR) index. In measurement instruments, the Cronbach’s α can either underestimate or overestimate scale reliability, depending on the amount of correlated measurement error. CR has the advantage of providing an estimate of scale reliability directly in the context of the CFA measurement model. It is recommended that the criterion for a reliable scale should be a CR > 0.7 (Brown, 2006). The convergent validity of the unidimensional constructs and the individual variables in the context of CFA tests the assumption that the variables correlate well with each other within their latent factor. Convergent validity was evaluated with the average variance extracted (AVE) that had to be above 0.5 and lower than the CR (Hair et al., 2010).

Discriminant validity within a first-order CFA analysis assumes that the different subscales of the SPSLI are distinctly different from one another. We evaluated this assumption with two different methods. The first method was to test a baseline multifactor model consisting of all of the instrument’s constituent subscales against a nested comparison model consisting of a single global factor. Evidence that the multifactor model fit the data significantly better than the one-factor model based on the χ² difference test supported discriminant validity of the multidimensional measurement model. The second method was to test whether different nested models, which allowed the correlation between pairs of constructs to be constrained to unity, and a base model, which allowed the correlations to be freed, were significantly different from each other. If the model with the freed correlations provided significant better fit based on the χ² difference test, discriminant validity was confirmed (Bryant, King, & Smart, 2007).

Testing the Hypothesis That the Responses to the SPSLI Can be Explained by Five First-Order Factor and One Second-Order Factor

In this part of the analysis, a second-order CFA model was tested, where the focus shifted to the correlations among the factors. The goal of this higher order factor analysis was to provide a more parsimonious account for the correlations among the five lower order factors explaining the higher order factor of love. Higher order CFA models are often used for theory testing as was the case in this analysis (Brown, 2006). We used the following order in our analysis to develop the higher order factor solution: (1) developed a well-behaved first-order CFA solution; (2) examined the magnitude and pattern of correlations among factors in the first-order solution; and (3) fitted the second-order factor model, as justified by Sorokin’s love theory. After the second-order model was established, we again calculated a CR and AVE for the second order factor named love.

Testing the Equivalence of the Measurement and Structural Models Across Males and Females

We then turned our focus to testing our focus to test whether the measurement and structural models of the first- and second-order CFA solutions were equivalent across males and females, using multigroup CFA (MCFA). We wanted to ensure that the newly developed SPSLI could reliably be used across gender groups and that observed differences between males and females reflected true differences in the amount or variability of the different subscales. Within scale development, this can be seen as a test of reliability across different samples. We accomplished this with an MCFA analysis on the first-order model. We then proceeded to investigate any structural invariance across gender groups to see how the different latent factors were distributed and related to males and females. Within scale development, this can be seen as a test of construct validity. We accomplished this also with MCFA analysis on both the first- and the second-order models (Byrne, 2010). Measurement invariance was evaluated with the more practical CFI difference approach, where a reduction of 0.01 in CFA between baseline and subsequent models was interpreted as the threshold for measurement invariance (Cheung & Rensvold, 2002).

Data Screening

Preliminary analysis indicated that 10 cases did not provide enough information to be valuable for analysis and were subsequently discarded from the study. Two more outlier cases were also removed, resulting in a final sample of 518. This final sample size was deemed adequate for the analysis. For the EFA analysis, a common rule of thumb is to ensure a person to item ratio of 10:1 (Tabachnick & Fidell (2012). With a sample size of 518 and a 20-item scale, the person to item ratio was 25.9:1. Schreiber, Stage, King, Nora, and Barlow (2006) suggest that at least 10 participants are needed per estimated parameter in a CFA analysis. In the first-order CFA analysis, we specified 10 regressions, 9 covariances, and 20 variances, totaling 39 parameters that needed to be estimated, resulting in an acceptable ratio of 13.3 participants to one parameter estimated. For the second-order CFA analysis, we specified 14 regressions and 21 variances, totaling 35 parameters that needed to be estimated, resulting in an acceptable ratio of 14.8 participants to one parameter estimated.

No missing values on any of the items were more than 10% of the overall values and they were random, therefore, they were replaced by the series mean (Tabachnick & Fidell, 2012). After the removal of the two multivariate outlier cases, all items had scores that were normally distributed and multivariate normality was confirmed. No problematic multicollinearity was detected.

Results

Sample Description

Sample demographics are summarized in Table 1. The sample was 71% female, 84% Caucasian, with 76% working full time and 33% earning over $75,000 annually. The age range of the sample was between 21 and 77 with a mean age of 46.33 (SD = 11.70), having on average two children. The sample was well educated with the mean years of education being 16.67 (SD = 3.44). Three quarters of the sample (75%) considered themselves spiritual, with only 34% who indicated that they attend religious services.

Table 1.

Demographics of Sample.

Variable	F	%
Gender: n = 518
Male	148	28.6
Female	370	71.4
Race: n = 512
Caucasian/European American	432	84.4
African American/Black	56	10.9
American Indian/Native American	5	1.0
Asian/Asian American	3	0.6
Hispanic/Latino/Latina	6	1.1
Other	10	2.0
Household income: n = 518
<$15,000	28	5.4
$15,000–$25,000	45	8.7
$25,000–$50,000	149	28.8
$50,000–$75,000	124	24.0
>$75,000	172	33.2
Employment status: n = 515
Working full time	393	76.3
Working part time	33	6.4
Unemployed	35	6.8
Retired	25	4.9
Student	24	4.7
Keep house	5	1.0
Current marital status: n = 513
Married	182	35.5
Divorced	198	38.6
Separated	110	21.4
Widowed	1	0.2
Member of an unmarried couple	22	4.3
Spiritual: n = 518
Yes	388	74.9
No	130	25.1
Attend religious services: n = 518
Yes	177	34.2
No	341	65.8
	M (SD)	Min–max
Age	46.33 (11.70)	21–77
Years of education	16.67 (3.44)	6–30
Number of children	1.85 (1.15)	0–6

Note. M = mean; SD = standard deviation.

Divorce-related demographics are summarized in Table 2. In regard to the divorce demographics, for 77% this was their first divorce, with 68% who initiated the divorce. Just over a third of the sample (39%) indicated that the main reason for the divorce was differences in priorities and expectation. A third of the sample (32%) was going through the divorce at the time of the study.

Table 2.

Divorce Related Demographics.

	f	%
Reason for divorce: n = 518
Infidelity	136	26.3
Abuse	123	23.7
Differences in priorities and expectations	200	38.6
Other	59	11.4
Time since the divorce: n = 518
Currently in the divorce process	165	31.9
Between 1 and 5 years	90	17.4
Between 6 and 15 years	133	25.7
More than 15 years	130	25.1
Times divorced: n = 518
0	45	8.7
1	397	76.6
2 or more	76	14.7
Times initiated the divorce: n = 506
0	153	30.2
1	314	62.1
2 or more	39	7.7
The person initiating divorce: n = 505
Yes	345	68.3
No	160	31.7

Reliability and Content Validity on the Item Level for Subscales as Unidimensional Constructs

EFA revealed reliable and valid unidimensional 3- or 4-item constructs for the five domains of love contained in the SPSLI. One item each was removed from the original 4-item models for intensity and purity. The reliability of the subscales ranged between 0.79 and 0.86, with duration being slightly lower than the 0.80 criterion. The IIC matrixes for all the different subscale items were above 0.30, all corrected ITC and factor loadings (FL) were at or above the 0.45 threshold, and the mean of all corrected ITCs were above 0.50 (Table 3).

Testing the Underlying Factor Structure With an EFA

The EFA focused on testing the assumption that love is a five-factor structure composed of intensity, extensity, purity, adequacy, and duration. A KMO value of 0.94 indicated the existence of factors in a correlation matrix, making the data applicable for EFA analysis. The oblique rotation was verified with the factor correlation matrix showing correlations between the factors above the 0.32 threshold.

The first EFA analysis revealed problems with cross loadings on 2 items, namely Item 7 (extensity): “Even though he/she is no longer part of my family, I am kind toward my former spouse” and Item 14 (adequacy): “When I know it is good for the well-being of my former spouse, I provide for him/her.” Also, Item 18 (duration) loaded stronger on purity than on duration: “I will always take actions to help my former spouse.” These 3 items were therefore deleted from the final EFA analysis.

The final results are shown in Table 4. From the pattern matrix and the item communalities after rotation, it is clear that the solution is a good fit, with no violation related to item loadings or cross loadings. One violation was detected with the rotated communalities, showing Item 16 with a low communality of 0.28. However, this item could not be removed due to the subscale duration having only 3 items left. The final solution explained 59.36% of the total variance and converged within six iterations.

Table 3.

Corrected Item-Total Correlations (ITC), Factor Loadings (FL), Reliability (Cronbach’s α) and Inter-Item Correlations (IIC) for SPSLI Subscales.

Item	Intensity	Cronbach’s α = .81	Mean IIC = 0.59	ITC	FL
1	I will lose something cherished to help my former spouse meet his or her needs			0.70	0.84
2	I will lose something cherished to provide for my former spouse’s well-being			0.67	0.78
3	I will sacrifice my own needs to meet the needs of my former spouse			0.60	0.67
			Mean	0.66	0.77
	Extensity	Cronbach’s α = .82 (0.79)^a	Mean IIC = 0.53 (0.55)
4	I choose to have compassion for my former spouse without judging his or her behaviors			0.71 (0.69)	0.82 (0.84)
5	I act sympathetic toward my former spouse			0.67 (0.65)	0.76 (0.77)
6	I want my former spouse to be happy despite what happened between us			0.59 (0.55)	0.65 (0.62)
7	Even though he or she is no longer part of my family, I am kind toward my former spouse			0.62	0.70
			Mean	0.65 (0.63)	0.73 (0.74)
	Purity	Cronbach’s α = .81	Mean IIC = 0.59
8	I give freely to my former spouse without expecting anything back			0.62	0.71
9	I provide for my former spouse’s well-being without expecting anything back			0.69	0.81
10	I support my former spouse without needing support back			0.67	0.78
			Mean	0.66	0.77
	Adequacy	Cronbach’s α = .86 (0.82)	Mean IIC = 0.62 (0.61)
11	When I know it will be healthy for my former spouse, I give what he or she needs			0.72 (0.68)	0.80 (0.79)
12	When I know it will result in something positive for my former spouse, I give what he or she needs			0.67 (0.65)	0.73 (0.74)
13	When I know my former spouse will benefit, I act with compassion toward him or her			0.73 (0.70)	0.81 (0.81)
14	When I know it is good for the well-being of my former spouse, I provide for him or her			0.73	0.80
			Mean ITC	0.71 (0.67)	0.78 (0.78)
	Duration	Cronbach’s α = .79 (0.71)	Mean IIC = 0.49 (0.45)
15	Every once in a while I help my former spouse			0.59 (0.55)	0.67 (0.70)
16	I put in time to develop a better relationship with my former spouse			0.45 (0.45)	0.50 (0.53)
17	I put in the energy to remain friends with my former spouse			0.71 (0.59)	0.86 (0.80)
18	I will always take actions to help my former spouse			0.66	0.78
			Mean ITC	0.60 (0.53)	0.70 (0.68)

^aValues in parenthesis are from repeating the subscale analysis after EFA resulted in the deletion of 3 items.

Note. EFA = exploratory factor analysis; SPSLI = Sorokin Psychosocial Love Inventory.

Table 4.

Pattern Matrix of EFA solution for SPSLI.

2.	Item	Intensity	Extensity	Purity	Adequacy	Duration	Communality
3.	1	0.84	0.00	−0.08	−0.02	0.09	0.68
4.	2	0.85	0.00	0.00	0.05	−0.13	0.66
5.	3	0.48	−0.06	0.31	−0.03	0.10	0.53
6.	4	0.00	0.79	0.13	0.03	−0.11	0.64
7.	5	0.04	0.58	0.00	0.26	0.01	0.72
8.	6	−0.04	0.73	−0.10	−0.08	0.08	0.41
9.	8	0.14	0.11	0.58	−0.20	0.20	0.58
10.	9	0.07	−0.08	0.68	0.31	−0.16	0.71
11.	10	−0.08	0.05	0.63	0.13	0.10	0.62
12.	11	0.09	0.10	0.09	0.46	0.18	0.62
13.	12	0.09	−0.02	−0.15	0.63	0.24	0.54
14.	13	−0.08	0.02	0.10	0.82	−0.01	0.74
15.	15	−0.06	−0.11	0.15	0.18	0.61	0.58
16.	16	0.02	0.01	−0.14	0.07	0.57	0.28
17.	17	−0.02	0.11	0.16	−0.05	0.62	0.59

Note. EFA = exploratory factor analysis; SPSLI = Sorokin Psychosocial Love Inventory.

The unidimensional analysis was then repeated to investigate the reliability and validity of the three constructs where items were deleted. The final reliability of the subscales ranged between 0.71 (duration) and 0.82 (adequacy). The IIC, corrected ITC, FL, as well as the mean of all corrected ITC met the required thresholds (Table 3 with revised analysis data in parentheses).

Confirming the EFA Solution With a First-Order CFA

The next step was to confirm the EFA solution with a first-order CFA. The results showed good fit with all fit indices above the criterion (Table 5) and statistically significant FL (Table 6). The results also indicated reliable subscales based on the CR index (Table 7), with coefficients being very similar to the Cronbach’s α coefficients. Convergent validity was confirmed with the AVEs above 0.50 for all but the duration subscale and the CR higher than the AVE for all subscales (Table 7). Discriminant validity was confirmed as indicated in Table 8. The χ² difference test indicated that the base multifactor model was a better fit than the single global factor model. Also, the different paired constraint models all showed worse fit than the base model.

Table 5.

Fit Indices for Multidimensional CFA Models.

	χ² (df)	p	χ²/df	GFI	CFI	RMSEA	SRMR
Multidimensional models
First order	242.48 (80)	0.00	3.03	0.94	0.96	0.06	0.04
Second order	274.38 (85)	0.00	3.23	0.94	0.95	0.07	0.05
Equivalence between males and females (first-order model)
Unconstrained multigroup model	343.85 (160)	0.00	2.15	0.92	0.951	0.05	0.05
Measurement weights invariant	363.48 (170)	0.00	2.14	0.92	0.948	0.05	0.05
Structural covariances invariant	380.34 (185)	0.00	2.06	0.91	0.948	0.05	0.05
Measurement residuals invariant	416.94 (200)	0.00	2.08	0.91	0.942	0.05	0.05
Factorial equivalence between males and females (second-order model)
Unconstrained multigroup model	386.87 (172)	0.00	2.25	0.91	0.943	0.05	0.07
Measurement weights invariant	404.59 (182)	0.00	2.22	0.91	0.941	0.05	0.07
Structural weights invariant	408.01 (186)	0.00	2.19	0.91	0.941	0.05	0.08
Structural covariances invariant	408.94 (187)	0.00	2.19	0.91	0.941	0.05	0.08
Structural residuals invariant	413.29 (191)	0.00	2.16	0.91	0.941	0.05	0.08
Measurement residuals invariant	449.51 (206)	0.00	2.18	0.90	0.935	0.05	0.08

Note. CFA = confirmatory factor analysis; CFI = comparative fit index; df = degrees of freedom; GFI = goodness-of-fit index; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual.

Criteria for good fit—χ²/df < 3; GFI ≥ 0.92; CFI ≥ 0.92; RMSEA < 0.06; SRMR < 0.08.

Table 6.

Standardized and Unstandardized Coefficients for the SPSLI as a First and Second-Order Multidimensional Scales.

		First-Order CFA			Second-Order CFA
Observed Items	Latent Construct	β	B	SE	β	B	SE
1	Intensity	0.81	1.00		0.81	1.00
2	Intensity	0.76	0.94	0.05	0.77	0.94	0.05
3	Intensity	0.74	0.90	0.06	0.73	0.89	0.06
4	Extensity	0.80	1.00		0.80	1.00
5	Extensity	0.83	1.06	0.06	0.84	1.07	0.06
6	Extensity	0.58	0.63	0.05	0.58	0.63	0.05
8	Purity	0.72	1.00		0.71	1.00
9	Purity	0.81	1.06	0.06	0.82	1.09	0.06
10	Purity	0.77	1.07	0.07	0.78	1.09	0.07
11	Adequacy	0.81	1.00		0.77	1.00
12	Adequacy	0.71	0.94	0.06	0.71	0.94	0.06
13	Adequacy	0.81	0.99	0.05	0.80	0.98	0.05
15	Duration	0.76	1.00		0.74	1.00
16	Duration	0.50	0.73	0.07	0.50	0.74	0.07
17	Duration	0.76	1.09	0.07	0.77	1.12	0.07

Note. CFA = confirmatory factor analysis; SE = standard error; SPSLI = Sorokin Psychosocial Love Inventory.

All estimates were significant at α = 0.05 level.

Table 7.

Composite Reliability and Average Variance Extracted.

	CR	AVE
First-order CFA
Intensity	0.81	0.59
Extensity	0.79	0.56
Purity	0.81	0.59
Adequacy	0.81	0.61
Duration	0.71	0.47
Second-order CFA
Love	0.87	0.63

Note. AVE = average variance extracted; CFA = confirmatory factor analysis; CR = composite reliability.

Table 8.

Discriminant Validity.

Model	χ2	df	χ² diff p
Base multifactor model	242.5	80
Single global factor model	755.9	90	0.001
Intensity–extensity equality constraints	335.7	81	0.001
Intensity–purity equality constraints	287.8	81	0.001
Intensity–adequacy equality constraints	318.0	81	0.001
Intensity–duration equality constraints	300.0	81	0.001
Extensity–purity equality constraints	292.6	81	0.001
Extensity–adequacy equality constraints	301.5	81	0.001
Extensity–duration equality constraints	307.0	81	0.001
Purity–adequacy equality constraints	286.2	81	0.001
Purity–duration equality constraints	275.0	81	0.001
Adequacy–duration equality constraints	283.5	81	0.001

Note . df = degrees of freedom.

Testing the Hypothesis That the Responses to the SPSLI Can be Explained by Five First-Order Factors and One Second-Order Factor

The second-order multidimensional model tested assumed that the five specific domains of love could be explained by a higher order love factor. These results also showed similar fit as the first-order CFA model (Table 5) and statistically significant FL (Table 6). The final second-order model, with all its standardized estimates is shown in Figure 1. As can be seen from the squared multiple correlations, the second-order love factor is best able to explain purity (96%) and adequacy (85%), followed by duration (74%). The second-order love factor was least able to explain extensity (63%) and intensity (46%). The CR for this final higher order love scale shows a reliable scale and the AVE shows a valid higher order construct (Table 7).

Figure 1.

Final second-order CFA model for the SPSLI with all its standardized estimates showing the squared multiple correlations of the five constructs. CFA = confirmatory factor analysis; SPSLI = Sorokin Psychosocial Love Inventory.

Testing the Equivalence of the Measurement and Structural Models Across Males and Females

MCFA revealed both measurement equivalence and structural equivalence between males and females in both the first- and the second-order models. As seen in Table 5, the CFI did not change beyond the 0.01 threshold between the models, indicating no significant change in fit as a result of adding additional equality constraints to the multigroup unconstrained model. Any differences between males and females could therefore be attributed to real differences in how the different gender groups were able to show love actions toward ex-spouses and not differences due to measurement or structural inequivalence.

Discussion

The primary focus of this study was to develop and validate the SPSLI for divorced individuals. We first used traditional psychometric techniques to test the internal consistency and validity of the instrument, after which we proceeded with an EFA to test the underlying factor structure of the SPSLI. We then completed a series of CFAs to confirm the factor structure and to provide additional evidence that the SPSLI can be used as a reliable and valid scale for measuring potential love actions portrayed toward former spouses during or after a divorce. All fit indices for the different CFA models were within the acceptable range. We also tested the equivalence of the multidimensional models between males and females with a series of MCFA analysis and found them to be both measurement and structurally invariant across gender groups.

Although this study showed promising results, the analysis done thus far was only on one sample to save costs and will have to be repeated with different samples before a reliable and valid SPSLI can be assumed. Also, the sample was not very diverse and consisted mainly of White, educated, middle class females, during various stages of the divorce process with close to a third of the sample currently going through a divorce in contrast to half of the sample being divorced for longer than 6 years. People currently going through a divorce may respond differently to a scale focused on love actions than those who have been divorced longer, which could result in measurement and structural invariances not tested with this study. In follow-up studies, more diverse samples should be recruited and an MCFA should be completed to test measurement and structural invariance between groups of individuals going through a divorce and groups of individuals divorced for a longer time.

Throughout the analysis, the duration subscale did not perform as well as the other subscales. The Cronbach’s α was below the recommended threshold of .80 at .71 and the convergent validity index (AVE) was below the recommended threshold of 0.50 at 0.47. Item 16 (I put in time to develop a better relationship with my former spouse) which formed part of this subscale did not perform well in any of the analysis and had consistently the lowest FL as well as a problematic rotated commonality value in the EFA analysis. In follow-up studies, we recommend new items be designed for this subscale and evaluated again within the context of the other subscales to develop a better performing unidimensional scale.

The second-order multidimensional model tested showed the second-order love factor to adequately explain most of the latent construct, except for intensity that was significantly lower than the rest of the constructs. It may be needed to carefully review this construct against Sorokin’s love theory to determine whether the items are an adequate reflection of what Sorokin meant when he described intensity. It may also be an indication that to love intensely with a willingness to sacrifice may be an ideal that is difficult for divorced individuals to reach.

Once the scale has been tested against a new sample, it will become important to evaluate the construct validity of the SPSLI against other measures of love that currently exists in the literature. We believe that Sorokin intended to describe something different from the current available instruments related to compassion, empathy, sympathy, and altruism. This assumption will have to be tested in subsequent studies.

This article adds to the literature that promotes love, care, and cooperation between individuals, especially divorced spouses. With the majority of divorced individuals being parents, Sorokin’s multidimensional theory of love allows a way to evaluate love actions in order to improve postdivorce relationships that could prove beneficial to children. The instrument can be used both in research studies and in clinical work to promote a better understanding of how people who divorce can access their intrinsic self to promote love actions toward others, even if it is a former spouse. Divorce is a fact of life that will continue to occur for married individuals. If it can be promoted as a positive step as has been documented in some literature (Hetherington, 2003; Masheter, 1998; Schneller & Arditti, 2004; Wallerstein & Kelly, 1980), rather than being represented as a negative step in life, love actions can become the focus of a new positive paradigm. After evaluating actions with the help of the SPSLI, the results gained may help divorcing individuals, their therapists, or mediators find ways to help shift potential angry behaviors to other-regarding love actions within a short time. This would alleviate the damage that anger causes to individuals, their families, friends, and society.

Also, having a scale that provides a framework to conceptualize love using five dimensions can guide divorcing individuals to know when they are loving intensely, even when they are losing something; loving extensively, in order to spread love; loving purely, not expecting reciprocity; loving adequately, aligning intent with another’s needs; and, being able to sustain love actions for a life time. These actions will not only make the process of divorce less tumultuous but also allow individuals to grow inwardly and develop the character needed in our society to promote love.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Amato

P. R.

(2000). The consequences of divorce for adults and children. Journal of Marriage and the Family, 62, 1269–1287.

Amato

P. R.

(2010). Research on divorce: Continuing trends and new developments. Journal of Marriage and Family, 72, 650–666.

Berscheid

(2010). Love in the fourth dimension. Annual Review of Psychology, 61, 1–25.

Bierman

Fazio

E. M.

Milkie

M. A.

(2006). A multifaceted approach to the mental health advantage of the married: Assessing how explanations vary by outcome measure and unmarried group. Journal of Family Issues, 27, 495–503.

Bonarch

Sales

(2002). Forgiveness as a mediator between post divorce cognitive processes and coparenting quality. Jounral of Divorce & Remarriage, 38, 17–38.

Booth

Amato

P. R.

(2001). Parental predivorce relations and offspring. Child Development, 63, 197–212.

Brotherson

S. E.

Duncan

W. C.

(2004). Rebinding the ties that bind: Government efforts to preserve and promote marriage. Family Relations, 53, 459–468.

Brown

T. A.

(2006). Confirmatory factor analysis for applied research. New York, NY: Guilford.

Bryant

F. B.

King

S. P.

Smart

C. M.

(2007). Multivariate statistical strategies for construct validitation in positive psychology. In Ong

A. D.

van Dulmen

M. H. M.

(Eds.), Oxford handbook of methods in positive psychology. New York, NY: Oxford University Press.

10.

Byrne

B. M.

(2010). Structural equation modeling with AMOS (2nd ed.). New York, NY: Taylor & Francis Group.

11.

Chatav

Whisman

M. A.

(2007). Marital dissolution and psychiatric disorders: An investigation of risk factors. Journal of Divorce & Remarriage, 47, 1–13.

12.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255.

13.

Chung

M. C.

Farmer

Grant

Newton

Payne

Perry

… Stone

(2003). Coping with post-traumatic stress symptoms following relationship dissolution. Stress and Health, 19, 27–36.

14.

Costello

A. B.

Osborne

J. W.

(2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10, 1–9.

15.

de Waal

F. B. M.

(2009). Putting altruism back into altruism: The evolution of empathy. Annual Review of Psychology, 59, 279–300.

16.

DeVellis

(2012). Scale development: Theory and application (3rd ed.). Los Angeles, CA: Sage.

17.

Duck

(2007). Human relationships (4th ed.). Thousand Oaks, CA: Sage.

18.

Faul

A. C.

van Zyl

M. A.

(2004). Constructing and validating a specific multi-item assessment evaluation. In Roberts

A. R.

Yeager

K. R.

(Eds.), Evidence based practice manual: Research and outcome measures in health and human services (pp. 564–584). New York, NY: Oxford University Press.

19.

Fawcett

Reynolds

(2010). Mental health and older women: The challenges for social perspectives and community capacity building. British Journal of Social Work, 40, 1488–1502.

20.

Fehr

(2006). A prototype approach to studying love. In Sternberg

R. J.

Weis

(Eds.), The new psychology of love (pp. 225–246). New Haven, CT: Yale University Press.

21.

Frisco

M. L.

Muller

Frank

(2007). Parents’ union dissolution and adolescents’ school performance: Comparing methodological approaches. Journal of Marriage and Family, 26, 3–31.

22.

Gerdes

K. E.

Lietz

C. A.

Segal

E. A.

(2011). Measuring empathy in the 21st century: Development of an empathy index rooted in social cognitive neuroscience and social justice. Social Work Research, 35, 83–93.

23.

Hair

Black

Babin

Anderson

(2010). Multivariate data analysis (7th ed.). Upper Saddle River, NJ: Prentice Hall.

24.

Hango

D. W.

Houseknecht

S. K.

(2005). Marital disruption and accidents/injuries among children. Journal of Family Issues, 26, 3–31.

25.

Hatfield

Sprecher

(1986). Measuring passionate love in intimate relationships. Journal of Adolescence, 9, 383–410.

26.

Hazan

Shaver

(1987). Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology, 52, 511–524.

27.

Hendrick

(1986). A theory and method of love. Journal of Peronality and Social Psychology, 50, 392–402.

28.

Hetherington

E. M.

(2003). Intimate pathways: Changing patterns in close personal relationships across time. Family Relations, 54, 318–331.

29.

L. T.

Bentler

P. M.

(1995). Evaluating model fit. In Hoyle

R. H.

(Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 76–99). Thousand Oaks, CA: Sage.

30.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.

31.

Hughes

M. H.

Waite

L. J.

(2009). Marital biography and health at mid-life. Journal of Health and Social Behavior, 50, 344–358.

32.

Huston

T. L.

Melz

(2004). The case for (promoting) marriage: The devil is in the details. Journal of Marriage & Family, 66, 943–958.

33.

IBM Corporation. (2011). IBM SPSS Amos 20.0.0. Meadville, PA: Amos Development Corporation.

34.

Johnson

(2004). The practice of emotionally focused marital therapy creating connetions (2nd ed.). New York, NY: Brunner-Routledge.

35.

Kelly

J. B.

Emery

R. E.

(2003). Children’s adjustment following divorce: Risk and resilience perspectives. Family Relations, 52, 352–362.

36.

Lasswell

T. E.

Lasswell

M. E.

(1976). I love you but I’m not in love with you. Journal of Marriage & Family Counseling, 38, 211–224.

37.

Levin

(2000). A prolegomenon to an epidemiology of love: Theory, measurement, and health outcomes. Journal of Social and Clinical Psychology, 19, 117–136.

38.

Levin

Kaplan

B. H.

(2010). The Sorokin multidimensional inventory of love experience (SMILE): Development, validation, and religious determinants. Review of Religious Research, 51, 380–401.

39.

Levite

Cohen

(2012). The tango of loving hate: Couple dynamics in high-conflict divorce. Clinical Social Work Journal, 40, 46–55.

40.

Linley

Joseph

Harrington

Wood

(2006). Positive psychology: Past, present and (possible) future. Journal of Positive Psychology, 1, 3–16.

41.

Marsh

H. W.

Hau

K. T.

Wen

(2004). In search of golden rules: Comment on hypothesis testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Structural Equation Modeling, 11, 320–341.

42.

Masheter

(1998). Divorce as selection, stress, and process. Journal of Divorce & Remarriage, 29, 144–159.

43.

Masuda

(2003). Meta-analyses of love scales: Do various love scales measure the same psychological constructs? Japanese Psychological Research, 45, 25–37.

44.

Netemeyer

R. G.

Bearden

W. O.

Sharma

(2003). Scaling procedures: Issues and applications. Thousand Oaks, CA: Sage.

45.

Nunnally

J. C.

Bernstein

I. H.

(1994). Psychometric theory (3rd ed.). New York, NY: McGraw-Hill.

46.

Peterson

(2006). A primer in positive psychology. New York, NY: Oxford University Press.

47.

Post

S. G.

(2003). Unlimited love: Altruism, compassion and service. Radnor, PA: Templeton Foundation Press.

48.

Potter

(2010). Psychosocial well-being and the relationship between divorce and children’s academic achievement. Journal of Marriage and Family, 72, 933–946.

49.

Rohde-Brown

Rudestam

K. E.

(2011). The role of forgiveness in divorce adjustment and the impact of affect. Journal of Divorce & Remarriage, 52, 109–124.

50.

Rubin

(1970). Measurement of romantic love. Journal of Peronality and Social Psychology, 16, 265–273.

51.

Saini

(2012). Reconceptualizing high-conflict divorce as a maladaptive adult attachment response. Famiilies in Society: The Journal of Contemporary Social Services, 93, 173–180.

52.

Saleeby

(2009). The strengths perspective in social work practice (5th ed.). Boston, MA: Allyn & Bacon.

53.

Schneller

D. P.

Arditti

J. A.

(2004). After the breakup: Interpreting divorce and rethinking intimacy. Journal of Divorce & Remarriage, 42, 1–37.

54.

Schreiber

J. B.

Stage

F. K.

King

Nora

Barlow

(2006). Reporting structural equation modeling and confirmatory factor analysis results: A review. The Journal of Educational Research, 99, 323–337.

55.

Sorokin

P. A.

(1950). Altruistic love. Boston, MA: The Beacon Press.

56.

Sorokin

P. A.

(1954a). Forms and techniques of altruistic and spiritual growth. Boston, MA: Beacon Press.

57.

Sorokin

P. A.

(1954b). The ways and power of love: Types, factors, and techniques of moral transformation. Philadelphia, PA: Templeton Foundation Press.

58.

Sorokin

P. A.

(1958). The reconstruction of humanity. Hartford, ME: Greenleaf Books.

59.

Sorokin

P. A.

(1963). A long journey. New Haven, CT: College and University Press.

60.

Sorokin

P. A.

(1965). Sociology of yesterday, today and tomorrow. American Sociological Review, 30, 833–843.

61.

SPSS. (2011). IBM SPSS for Windows. Armonk, NY: IBM.

62.

Sun

(2002). Children’s well-being during parent’s marital disruption process: A pooled time-series analysis. Journal of Marriage and Family, 64, 742–762.

63.

Tabachnick

B. G.

Fidell

L. S.

(2012). Using multivariate statistics (6th ed.). Needham Heights, MA: Allyn & Bacon.

64.

Troxel

W. M.

Matthews

K. A.

(2004). What are the costs of marital conflict and dissolution to children’s physical health. Clinical Child and Family Psychology Review, 7, 29–57.

65.

Underwood

L. G.

(2009). Compassionate love: A framework for research. In Fehr

Sprecher

Underwood

L. G.

(Eds.), The science of compassionate love (pp. 3–26). Malden, MA: Wiley-Blackwell.

66.

Waite

L. J.

Luo

Lewin

A. C.

(2009). Marital happiness and marital stability: Consequences for psychological well-being. Social Science Research, 38, 201–212.

67.

Wallerstein

J. S.

(2005). Growing up in the divorced family. Clinical Social Work Journal, 33, 401–418.

68.

Wallerstein

J. S.

Kelly

J. B.

(1980). Surviving the breakup: How children and parents cope with divorce. New York, NY: Basic Books.

69.

Wheaton

Muthen

Alwin

D. F.

Summers

G. F.

(1977). Assessing reliability and stability in panel models. In Heise

D. R.

(Ed.), Sociological methodology (pp. 84–136). San Francisco, CA: Jossey-Bass.

70.

Wilcox

W. B.

Marquardt

Popenoe

Whitehead

B. D.

(Eds.). (2010). The state of our unions: When marriage disappears: The new middle America. Charlottesville, VA: The National Marriage Project, Institute for American Values, University of Virginia. Retrieved from http://stateofourunions.org/

71.

Williams

Umberson

(2004). Marital status, marital transitions, and health: A gendered life course perspective. Journal of Health and Social Behavior, 45, 81–98.

72.

Worthington

R. L.

Whittaker

T. A.

(2006). Scale development research: A content analysis and recommendations for best practices. The Counseling Psychologist, 34, 806–838.

73.

Zhang

Hayward

(2006). Gender, the marital life course, and cardiovascular disease in late midlife. Journal of Marriage and Family, 68, 639–657.