Validity of the Breakup Strategies Questionnaire in Korea: Exploratory Structural Equation Modeling and Rasch's Modeling

Abstract

This study examines the validity of the Korean implementation of the Breakup Strategies Questionnaire. In Study 1, the construct validity of the Breakup Strategies Questionnaire was explored with 292 Korean undergraduate students. Exploratory Structural Equation Modeling and Rasch's Modeling suggest 30 items with four factors (Avoidance/withdrawal, Positive tone/self-blame, Open confrontation, and Indirect manipulation) and five points in response. Study 2 with 174 Korean undergraduate students suggested potential gender differences in breakup strategy dimensions and provided support for concurrent validity with a measure of personal styles of coping as well as discriminant validity with social desirability. Cultural implications for the use of breakup strategies are discussed.

Keywords

Mental health individualism/collectivism translated questionnaires applications of statistical technique factor analysis socially desirable responding validity of assessment romance/courting behavior

Introduction

Romantic breakup is typically accompanied by anger and sadness (Sbarra, 2006). A breakup experience is reported as a painful experience particularly for college students (Simon & Barrett, 2010), and approximately 70% of Korean college students who come to counseling expressed difficulties in personal relationships including romantic relationships (Chang, 2013; Park, 2015). Recently, violent responses have been reported in South Korea such as murdering or spraying hydrochloric acid on the partner who terminated the relationship (Kang, 2016). In a recent survey of Korean singles in their 20s and 30s (LSinfo, 2015), 46% replied that they have “anger management issues”. One of the stated causes of their anger was reported as “conflict between couples”; 15.3% of male participants and 24.4% of female ones. Moreover, 6.7% answered “romantic partner” to the item “I know someone who I suspect of having an anger management disorder.” Studying breakup strategies is significant not only because it influences others' responses (Banks, Altendorf, Greene, & Cody, 1987) but also because the person who is initiating the breakup may be coping with perceived physical or other dangers.

Breakup strategy is defined as individuals' verbal and non-verbal approach to terminating romantic relationships (Baxter, 1982). Cody (1982) and Baxter (1982, 1984) suggested that those who want to terminate romantic relationships use various breakup strategies because they have different levels of concern for their former partners after the breakup. Directness in breakup strategies tends to differ by the amount of care or concern expressed toward the recipient (Baxter, 1985; Sprecher, Zimmerman, & Abrahams, 2010). Literature suggests that indirect breakup strategies (e.g., Avoidance/withdrawal) are related to negative post-breakup outcomes (Metts, Cupach, & Bejlovec, 1989; Sprecher et al., 2010); in contrast, direct breakup strategies (e.g., Confrontation) are associated with fewer negative outcomes (Banks et al., 1987; Sprecher et al., 2010).

Some of the major scales for quantifying breakup strategies are Breakup or Disengagement strategies (Cody, 1982) and Baxter Breakup Strategies Questionnaire (Baxter, 1982, 1984). However, these tools do not include new breakup strategies mediated through new technologies (i.e., indirect breakup through social network services) and use outdated factor analytic methods (i.e., principal components analysis based on the Kaiser criterion of eigenvalue greater than one).

To address these limitations, Collins and Gillath (2012) revised Baxter's measure and developed a new Breakup Strategies Questionnaire (BSQ). Items which specifically discuss the termination of a romantic relationship were added (e.g., Start dating someone else in the hopes my partner would learn about my desire to end the relationship through my actions) as well as six items regarding new breakup strategies using social network services. As a result, there now are seven sub-factors of the BSQ. The seven breakup strategies include (a) Avoidance/Withdrawal (disclose little about my personal activities and interests whenever we talk), (b) Positive tone/Self-blame (take “blame” to try to prevent us from leaving on a “sour note” with one another), (c) Open confrontation (honestly convey my wishes to my partner), (d) Manipulation (ask a third party to inform my partner of my desire to terminate the relationship), (e) Cost escalation (become unpleasant to my partner in the hopes that s/he would make the first move), (f) Distant/Mediated communication (terminate the relationship without letting my partner know about it directly, by changing my relationship status on social media such as Facebook, Myspace, Friendster, or other social hubs), and (g) De-escalation (“waited it out” until conditions were conductive to breakup, e.g., until vacation time).

Although Collins and Gillath (2012)’s revised version of the BSQ showed evident improvement from that of Baxter (1982, 1984), some limitations still exist. From a psychometric properties standpoint, the BSQ's factor analytic process limits item selection in order to secure factor stability. Items such as “I blocked my partner from seeing me on Instant Messenger” (factor loading of .25) and “I dropped subtle ‘hints’ that things had changed between us” (factor loading of .26) were still used despite low factor loading. Thus, the factor structure which includes them is likely to be unstable. For this reason, this study uses a strict item retention standard, eliminating all items with factor loading <.40 (Comrey & Lee, 1992). We utilize more effective analysis strategies: the Exploratory Structural Equation Modeling (ESEM; Asparouhov & Muthén, 2009), which derives a realistic and understandable construct by verifying the correlation between items that measure different concepts, and the Rasch Model (Andrich, 1978; Rasch, 1960), which allows a test of individual item quality.

Second, the BSQ was developed from an (individualistic) Western context which could bias results through implicit assumptions that are not appropriate for South Korean (collectivistic) culture. For individualistic cultures, self-assertive breakup strategies could be natural since openly direct confrontation is common (Ting-Toomey, et al., 2000). By contrast, for collectivistic cultures such as South Korea, interpersonal harmony is emphasized (Choi, Yoon, Hahn, Cho, & Lee, 1999). Thus, announcing the breakup itself is even more difficult, and indirect breakup strategies (i.e., accommodating, avoiding styles, or involving third parties) could be preferred (Kozan, 1997).

From another angle, South Korea's smart phone usage is over 91% for the relevant population (KT Economy & Management Institute, 2016). We would expect a more technologically oriented set of breakup strategies; for instance, 9 out of 10 Koreans in their 20s–30s are reported to have broken up using smart phone instant messenger (Lim, 2016). Thus, it is necessary to validate whether the BSQ and its sub-factors developed in Western contexts 30 years ago are applicable for Korean college students today.

To test these ideas, the current study, consisting of two sub-studies, first translated the BSQ into Korean through a robust translation-back translation procedure (Sperber, Devellis, & Boehlecke, 1994), with the developer's permission. In Study 1, the construct validity of the Korean BSQ (BSQ-K) was examined by utilizing ESEM (which addresses the limitations of applying exploratory factor analysis first and then conducting confirmatory factor analysis). In addition, the overall individual item fitness of the BSQ-K was examined in detail by using Rasch's modeling. Study 2 sought to provide additional validity estimates of the BSQ-K by testing potential gender differences on the scores of the four sub-scales of the BSQ-K and its correlations with the subscales of the problem-focused style of coping (Reflective, Suppressive, and Reactive) for testing concurrent validity, and social desirability to test discriminant validity.

The hypotheses of the current study through two sub-studies are as follows.

Hypothesis 1: The factor structure of the Korean version of the BSQ (BSQ-K) will differ from that of the original scale (BSQ).

Hypothesis 2: There will be significant correlations among the sub-factors of the BSQ-K.

Hypothesis 3: The BSQ-K's sub-factors will differ by gender.

Hypothesis 4: There will be significant correlations between breakup strategies and problem-focused coping strategies.

Hypothesis 5: There will be no correlation between breakup strategies and social desirability.

Study 1

Method

Participants and procedure

Two hundred and ninety-two Korean college students (159 men, 133 women, mean age of 21.86, SD = 2.20; range 18–34 years) attending a private university in Busan were recruited as participants. Of the participants, the mean age of men was 22.68 (SD = 2.05; range 18–27 years) and women was 20.89 (SD = 1.98; range 18–34 years). Among them, 120 students were currently in a romantic relationship (41.1%), 160 students were single (52.8%), and average number of romantic relationships was 2.30 (SD = 1.62). Out of 154 men, 67 students (43.5%) were in a romantic relationship and 87 students (56.5%) were single. Of the 126 women, 53 students (42.1%) were in a romantic relationship and 73 students (57.9%) were single. The average duration of present or most recently terminated romantic relationship of the participants was reported at 11.79 months (SD = 13.10; range 0–75 months). The average duration of being in a romantic relationship for men were 12.22 months (SD = 22.51; range 0–75 months) and 11.15 months (SD = 13.48; range 0–72 months) for women. While their average time period passed after breakup was 17.68 months (SD = 22.51; range 0–217 months), men reported at 18.95 months (SD = 20.52; range 0–150 months) and women at 16.28 (SD = 24.72; range 0–217 months). The mean age of the participants at the time of most recent breakup was 20.37 years (SD = 3.88; range 13–33 years). Men reported at 20.96 years (SD = 4.34; range 15–28 years) and women reported at 19.61 years (SD = 3.10; range 13–33 years). In regards to their most recent terminated romantic relationship, 56 students (19.2%) replied that their partners terminated the relationship. By contrast, 93 students (31.8%) replied that they initiated termination of the relationship. Lastly, 100 students (34.2%) mutually agreed to break up. For men, 37 students (23.3%) replied that their partners terminated the relationship, 46 students (28.9%) terminated the relationship themselves, and 56 students (35.2%) mutually agreed to break up. As for women, 19 students (14.4%) were asked to breakup by their partners, 47 students (35.6%) initiated the breakup, and 44 students (33.3%) mutually agreed to breakup. The remainder of students had never gone through a breakup responded to the questionnaire hypothetically. The hard copy of the questionnaire was distributed during part of a class period for approximately 15 minutes. All participants read and agreed to the participant rights prior to answering the questionnaire, and all were given debriefing information regarding the research purpose. Although the institution had not established an official IRB procedure at the time of data collection, informed consents of research participation were obtained from the participants. An opportunity to raise any concerns or questions regarding the survey was provided. Participants were ensured that they could withdraw from participation at any time during the survey.

Measure

Breakup Strategies Questionnaire

The questionnaire which quantifies individual breakup strategies, the BSQ originally developed by Baxter (1982), modified by Sprecher et al. (2010), and then modified further by Collins and Gillath, (2012), was translated by the authors into Korean with the Collins' permission. The BSQ consists of seven breakup strategies with a total of 43 items. The seven breakup strategies include (a) Avoidance/Withdrawal, (b) Positive tone/Self-blame, (c) Open confrontation, (d) Manipulation, (e) Cost escalation, (f) Distant/Mediated communication, and (g) De-escalation, and details of these strategies were already provided in the Introduction section. Higher scores indicate more frequent use of a given breakup strategy. The responses were rated on a 7-point Likert-type scale (1 = very unlikely to 7 = very likely). Internal consistency reliability coefficient (Cronbach's α) among the BSQ items reported in Collins and Gillath (2012)'s research ranged from .60 (De-escalation) to .90 (Avoidance/Withdrawal).

Analysis

The BSQ-K was tested with the ESEM (Asparouhov & Muthén, 2009) and Rasch's Rating Scale Model (RSM) developed by Andrich (1978).

ESEM

The ESEM was applied to grasp the internal structure of the BSQ-K. ESEM is a combined technique of exploratory and confirmatory factor analysis; the combination is more useful for detailing a realistic internal structure than the traditional closed method (Asparouhov & Muthén, 2009). ESEM allows estimation and constraint of measurement errors between individual items, overcoming the problem of an unrealistic restriction of the correlation/covariance between each item's measurement error to 0 (Asparouhov & Muthén, 2009). Similarly, ESEM allows for the possibility that each item contributes toward multiple concepts (Marsh et al., 2009). The current study used Mplus 6.0 (Muthén & Muthén, 2010) for ESEM with the Maximum Likelihood estimation with Robust standard errors (MLR) for parameter estimation. In addition, common factor analysis rather than principal components analysis was used for the factor extraction method, and for a factor-rotation method, GEOMIN (a type of oblique rotation) was applied, in light of the high possibility of existing correlations among factors (Browne, 2001; McDonald, 2005). In order to evaluate the fitness of the exploratory factor analysis model, the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), RMSEA (Root Mean Square Error of Approximation), and SRMR (Standardized Root Mean Square Residual (SRMR) were used. Goodness of fit for the underlying model is considered desirable if both CFI and TLI are over .90 (Hong, 2000), RMESEA is under .06, and SRMR is under .08 (Hu & Bentler, 1999).

Item response theory

Although factor analysis can verify the structure of a scale, it cannot provide information on the appropriateness of individual items. Moreover, the use of inter-item correlation matrices between items may change the factor structure depending on the sample population's characteristics (Hong & Wong, 2005). Conversely, Item Response Theory (IRT) verifies the item appropriateness without being influenced by the sample population's characteristics. Thus, it increases the effectiveness of a scale, simultaneously evaluating the appropriateness of the item unit, thereby refining the scale (Fox & Jones, 1998; Heppner, Wampold, Owen, Thompson, & Wang, 2015). For this reason, this study utilized the Rasch model of IRT in order to evaluate the fitness of each item of the BSQ-K. The implementation selected was Andrich's (1978) RSM, Polytomous Model.

The equation for RSM (1978) is

\begin{matrix} \Pr {X_{ni} = x} = \frac{\exp \sum_{k = 0}^{x} (β_{n} (δ_{i} - τ_{k}))}{\sum_{j = 0}^{m} \exp \sum_{k = 0}^{j} (β_{n} - (δ_{i} - τ_{k}))} \end{matrix}

where δ_i means the difficulty of item i and τ_k indicates the kth threshold location of the rating scale which is in common to all the items. m denotes the maximum score and is identical for all the items. τ₀ is selected for computational convenience.

The Rasch model stably measures parameters of item difficulty and respondents' attributes even with relatively few items. In addition, the respondents' ability and item difficulty estimates can be compared mutually and independently placed on mutual measurement called logits. Since the Rasch model, unlike other item response theory models, satisfies specific objectivity, the respondents' ability and item difficulty may be independently and objectively estimated (Anderson, 1973). It is adequate in providing not only item characteristics but also information regarding fitness of response range numbers (Hong, Kim, & Wolfe, 2005). This study utilized WINSEPTS (Linacre, 2005) and evaluated the item fitness through a subject × item distribution chart according to the Rasch model, Infit Mean Square (item fitness and category probability curve on the respondents' category), and Outfit Mean Square.

Lastly, correlation analysis was performed to evaluate the association among sub-factors of the BSQ-K.

Results

Factor analysis

The aforementioned ESEM analysis was conducted in three rounds. In the first Exploratory Factor Analysis (EFA), exploratory rotation was conducted with reference to a tentative basic factor structure to discover a final solution which was then interpreted. In this process, as a robustness check, parallel analysis was used to examine a number of tentative factors in the BSQ-K. Parallel analysis is a precise method of factor retention (Velicer, Easton, & Fava, 2000). The aforementioned MLR was applied; this estimation method is robust to violations of measurement variables' normality and independency in factor structure. Any items whose factor loading measured less than .40 were deleted, and the resulting item content was carefully compared with the original scale.

Another EFA was performed excluding the items deleted by the first-round analysis. This yielded a superior fit; however, some items did not fit into any factors. As a result, in the third EFA analysis, unmatched items that had both high factor loading and cross factor loading (greater than .30) regarding the relative factor were estimated as free parameters. The results were then run through partially specified target rotation (Browne, 1972) which restrains the remaining uninterpreted factor loading to 0. In case of simultaneous significant correlation between measurement errors, relative measurement error covariance was estimated based on Modification Indices (MI; Byrne, 2001) prior to the analysis.

Before the ESEM analysis, parallel analysis was performed using FACTOR 10.4 (Lorenzo-Seva & Frrando, 2016) to examine a number of tentative factors in the Korean data. The results suggested four factors, unlike the Collins and Gillath's (2012) study. The fitness of the model suggested with the four-factor BSQ-K (total 43 items) was not strong, χ²(737) = 1347.94, CFI = .849, TLI = .815, RMSEA = .053, and SRMR = .047. The first factor consisted of 15 items (eigenvalue = 9.05): Avoidance/withdrawal (11 items), Distant/mediated communication (2 items; #7 & 23), De-escalation (1 item; #24), and Positive tone/self-blame (1 item; #12). The factor name was retained as Avoidance/withdrawal, since most of the items were also included in the original BSQ, and the remaining item content was also similar. The second factor has a total of 16 items (eigenvalue = 4.75): Manipulation (5 items; #15, 17, 19, 38, & 43), De-escalation (4 items; #18, 20, 35, & 36), Cost escalation (3 items; #14, 16, & 37), Distant/mediated communication (2 items; #26 & 28), and Positive tone/self-blame (2 items; #31 & 34) in the original BSQ. The factor loading on Manipulation was significantly higher than others. This factor was renamed as “Indirect manipulation,” since the overall content of the items pertained to indirectly expressing the desire to break up to intentionally make their partner initiate termination of the relationship. The third factor has a total of five items labeled as “Open confrontation” like that of the original BSQ including (eigenvalue = 3.08): Open confrontation (4 items; #1, 10, 39, & 42), and Cost escalation (1 item; #13). Finally, the fourth factor consists of a total of eight items labeled as “Positive tone/self-blame,” in concurrence with the item content in the original BSQ (eigenvalue = 2.06).

While examining the items within the four-factor structure, factor loading was examined with the purpose of deleting items with low factor loading during the first EFA to obtain a stable internal factor structure. To avoid low statistical power occurrence, the current study set a stricter standard than the previous work and deleted items under .40 factor loading. As a result, a total of 12 items were deleted: Avoidance/withdrawal (4 items; #2, 12, 23, and 24), Indirect manipulation (4 items; #20, 28, 29, and 34), Open confrontation (1 item; #13), and Positive tone/self-blame (3 items; #11, 22, & 31). As a result, the items well-supported per factor yielded a total of 31: Avoidance/withdrawal (11 items), Indirect manipulation (12 items), Open confrontation (4 items), and Positive tone/self-blame (4 items).

Next, the second EFA was processed on 31 items excluding the 12 items deleted through the first-round EFA. The fitness indices were CFI = .898, TLI = .864, RMSEA = .057, and SRMR = .041. The goodness of fit increased in comparison to the first EFA results and was adequate. The factor loadings of items per each factor yielded in the second EFA were high, all over .40 except for two items (#18 (.38) and 36 (.35)). Then, four possible items to perform cross factor loading (items over .30) for the third partially specified target rotation were examined: items #16 (.33), 18 (.34), 27 (.37), and 41 (.30). Since the cross factor loading of these items were impossible to interpret, they were estimated as free estimators in the third EFA. The remaining cross factor loads were restricted to 0.

Finally, in the third EFA, a partially specified target rotation was processed based on the results of the second EFA (see Table 1). Simultaneously, error covariance was connected if items over 10 MI showed covariance between measurement errors; this detected five pairs of adjacent items (#8 & 9, #5 & 4, #32 & 33, #35 & 36, and #40 & 41), and two pairs of items under the same factor in the original BSQ (#14 & 16 and #15 & 17). The fit indices were CFI = .937, TLI = .914, RMSEA = .045, and SRMR = .037, which was found most adequate. The factor structure shown through the third EFA was congruent to that of the BSQ suggested by Baxter (1982). In comparison to the confirmatory actor analysis result of 31 items (CFI = .797, TLI = .780, RMSEA = .072, and SRMR = .103), the ESEM analysis was confirmed to be adequate and more appropriate, considering practicality and possibility of interpretation.

Table 1.

The final (third) exploratory factor analyses (EFA) factor loading results and descriptive statistics.

Item	M	(SD)	F1	F2	F3	F4
F1 (Avoidance/withdrawal)
3. (Keep our conversations brief whenever we talk)	3.44	(1.88)	.73	−.02	.01	−.18
4. (Avoid scheduling future meetings with my partner whenever possible)	3.53	(2.00)	.81	.04	.10	−.07
5. (Maintain our conversations on a superficial level)	3.79	(1.78)	.81	−.05	−.05	.00
6. (Refrain from asking favors of my partner)	4.32	(1.90)	.78	−.15	.00	.04
7. (Terminate the relationship indirectly (through e-mail, text-messaging, or other unidirectional methods of communication))	2.82	(1.78)	.48	.20	−.17	−.20
8. (Disclose little about my personal activities and interests whenever we talk)	3.37	(1.86)	.84	−.03	−.04	.03
9. (Subtly discourage my partner from sharing aspects of his/her personal life with me)	3.74	(1.93)	.84	−.09	−.01	.01
25. (Devote more time to other people and activities)	4.54	(1.76)	.63	−.11	−.01	.10
30. (Reduce overt displays of liking and affection toward my partner)	4.43	(1.76)	.65	−.01	.14	.11
40. (Avoid contact with my partner as much as possible)	3.44	(1.74)	.55	.26	.07	−.08
41. (Cease doing favors for my partner)	3.62	(1.63)	.55	.25	.24	.06
F2 (Indirect manipulation)
14. (Pick an argument with my partner as an excuse to end the relationship)	2.64	(1.58)	.20	.50	.06	−.17
15. (Ask a third party (i.e., another person) to inform my partner of my desire to terminate the relationship)	1.67	(1.10)	−.13	.71	−.09	−.04
16. (Become unpleasant to my partner in the hopes that s/he would make the first move)	2.81	(1.77)	.32	.41	−.17	.04
17. (Intentionally “leak” my desire to end the relationship to someone I anticipate would inform my partner)	1.98	(1.35)	−.02	.67	−.05	−.04
18. (“Ease into” the ending of the relationship by saying it is just a “temporary thing”)	3.38	(1.78)	.34	.38	.01	.06
19. (Promote new relationships for my partner to make the ending of the relationship easier)	1.95	(1.32)	−.07	.71	−.06	.15
26. (Terminate the relationship without letting my partner know about it directly, by changing my relationship status on my webpage (facebook, myspace, Friendster, other webpages))	1.82	(1.31)	−.11	.60	−.07	−.12
35. (Try to find reasons for the end the relationship other than things about our relationship (e.g., a job offer, graduation, etc.))	3.16	(1.80)	.22	.48	.13	.20
36. (“Wait it out” until conditions were conducive to end the relationship (e.g., until vacation time))	3.02	(1.79)	.26	.34	−.03	.22
37. (Make the relationship more costly for my partner by being bitchy, demanding, etc.)	2.19	(1.41)	.05	.61	−.03	−.06
38. (Give hints of my desire to end the relationship to people who know the other person)	2.57	(1.69)	.09	.57	.12	−.03
43. (Start dating someone else in the hopes my partner would learn about my desire to end the relationship through my actions)	1.93	(1.39)	−.02	.50	.07	−.16
F3 (Open confrontation)
1. (Openly express to my partner my desire to end the relationship)	4.30	(1.83)	.03	.01	.61	−.24
10. (Verbally explain to my partner my reasons for desiring to end the relationship)	4.99	(1.62)	.02	−.07	.64	−.03
39. (Find a time and place when we can talk face to face about my desire to end the relationship)	4.36	(1.73)	.07	.12	.51	.29
42. (Honestly convey my wishes to my partner)	4.99	(1.44)	−.15	−.17	.63	.01
F4 (Positive tone/self-blame)
21. (Avoid hurting my partner's feelings at all costs)	4.44	(1.70)	−.07	−.04	−.19	.59
27. (Try to put my partner in a “good frame of mind” before breaking the news to him/her)	3.06	(1.88)	−.23	.45	.08	.60
32. (Try to prevent my partner from having any “hard feelings” about the end the relationship)	4.69	(1.61)	.07	.06	.14	.56
33. (Try to prevent us leaving on a “sour note” with one another)	5.19	(1.49)	.11	−.11	.06	.58

Note. The item numbers correspond to the original version of the Breakup Strategies Questionnaire (Collins & Gillath, 2012). F: Factor.

Rasch analyses

To confirm the fitness of each item Andrich's (1978) RSM of the Rasch model was processed for each of the four factors. The appropriateness of each item response category was tested using Hong et al.'s (2005) procedure. Inadequate items based on each item's fitness were deleted, followed by comparative analysis between the participants' characteristics pertaining to each breakup strategy and the degree of difficulty of agreement to each item based on the Person-item-map.

The appropriateness of response category

The category probability curve of each factor was examined to confirm applicability of the seven-factor structure of the BSQ; from 1 = very unlikely to 7 = very likely. In Figure 1, the X-axis represents the person measure relative to item difficulty of each breakup strategy. The Y-axis is the category probability, where each curve in the figure represents the response category from one to seven points. Each response category must show significantly high probability of selection in a particular area of the X-axis to be considered valuable. Mostly, items 2 and 3 on the response curve were not included in the response curve of 1 and 4. Thus, a 5-point Likert response category was confirmed to be more appropriate compared to a 7-point response category.

Figure 1.

Category probability curves of breakup strategy.

Item fit

Individual item validity was confirmed through Infit Mean Square (Infit MNSQ) and Outfit Mean Square (Outfit MNSQ). The Infit MNSQ and Outfit MNSQ validate the respondent or the item fitness by comparing the probability expected by the model and the subject's actual response rate, to confirm the appropriateness of the fit according to the Rasch's Model. The Outfit MNSQ is sensitive to abnormal item responses such as the respondent answering incorrectly, associated with the respondent's ability to answer a difficult item correctly. The Infit MNSQ by itself can unreasonably conclude that a fit is inappropriate due to few abnormal item responses. Hence, it is the appropriateness weighted on the abnormal item response to the items which are not highly beyond the subject's ability (Wright & Masters, 1982), which is desired. The expected result of the two numbers regarding satisfactory items is 1. Appropriateness was deemed unsatisfactory when the two numbers are simultaneously less than .60 or larger than 1.4 (Linacre, 2005; Wright, Linacre, Gustafson, & Martin-Lof, 1994). In case of Avoidance/withdrawal, the item fitness for item #7 (“(I would) Terminate the relationship indirectly (through e-mail, text-messaging, or other unidirectional methods of communication)”) was unsatisfactory (Infit MNSQ = 1.70, Outfit MNSQ = 2.07). On the other hand, the remaining 10 items showed appropriate item fitness. The items of the remaining three factors also showed appropriate fitness. Therefore, item #7 ((Terminate the relationship indirectly (through e-mail, text-messaging, or other unidirectional methods of communication)) was removed from the final breakup strategies.

Person-item-map

The participants' attributes for each breakup strategy and item difficulty were examined through a Person-item-map (see Figure 2). This graph transferred these aspects into logit, to directly compare the two numbers. In each map in Figure 2, the respondent's attributes are recorded on the left of the horizontal line in the middle, and item difficulty distribution is represented on the right. When the two distribution ranges match, each item difficulty is appropriately distributed to measure all ranges of breakup strategies attributes (Hong et al., 2005).

Figure 2.

Distribution of persons and items.

Items of the Avoidance/withdrawal factor showed equal and even distribution with the respondents' Avoidance/withdrawal attributes. By contrast, the items for Indirect manipulation showed a low distribution of the Indirect manipulation attribute compared to the item difficulty distribution. Thus, the Indirect manipulation attributes of the participants themselves are low. Such aspect may cause difficulty in measuring the respondents with Indirect manipulation attributes. Items of Open confrontation showed overall average level item difficulty, while items measuring high Open confrontation attribute responses were lacking. Item difficulty of Positive tone/self-blame items showed wide distribution despite the small number (4) of items.

In conclusion, the BSQ-K was confirmed to have a four-factor structure (30 items) through the ESEM and Rasch analysis. The total item and alpha coefficients (internal consistency estimates among items) for each factor are Avoidance/withdrawal (10 items; α = .92), Indirect manipulation (12 items; α = .85), Open confrontation (4 items; α = .68), and Positive tone/self-blame (4 items; α = .71). Thus, Hypothesis 1 (the BSQ-K's factor structure will differ from that of the original BSQ) was supported.

Correlations among the BSQ-K sub-factors

Correlation analysis was performed to examine the association among the four breakup strategies. The correlations are as follows: Avoidance/withdrawal showed a positive correlation with Indirect manipulation (r = .39, p < .001) and Open confrontation (r = .21, p < .001); and Open confrontation and Positive tone/self-blame (r = .14, p = .020) showed a significant positive correlation. Thus, Hypothesis 2 (there will be significant correlations among the breakup strategies) were partially supported.

Discussion

In Study 1, the seven-factor structure of the modified BSQ (Collins & Gillath, 2012) was examined and validated with a sample of Korean college students through ESEM. Unlike the original seven-factor (43 items) BSQ model, in the Korean sample, a four-factor structure (30 items) was suggested, which consisted of (a) Avoidance/withdrawal, (b) Positive tone/self-blame, (c) Open confrontation, and (d) Indirect manipulation. While three of the seven factors in the original model (Avoidance/withdrawal, Positive tone/self-blame, and Open confrontation) remained the same in the Korean sample, the remaining four factors (Cost escalation, Manipulation, Distant/Mediated communication, and De-escalation) were discovered to be most accurately described by one combined factor, Indirect manipulation. Interestingly, the four-factor structure corresponds with the original model of the BSQ suggested by Baxter (1982).

Research results derived from Rasch's Rating Scale Modeling are significant in that they provided understanding of the BSQ-K's characteristics not only from a sub-factor perspective but also from the level of detailed items. For example, after comparing participants' ability and item difficulty through the Rasch Model's Person-item map, said ability showed a low distribution in contrast to the Indirect manipulation items measuring average level item difficulty. In other words, not only did few people responded as using Indirect manipulation, but the center point of the distribution regarding item difficulty was not congruent with the range where methods of manipulation are measurable. Such an outcome may reflect the fact that causing emotional harm to others to initiate termination of a relationship is even more unapproved of in Korean society. One of the hypotheses was that the BSQ-K might still express implicit biases from the fact that it was derived from the BSQ and that these biases might make it a poor measure. In fact, the hypothesized bias (toward indirect means of breaking off relationships) ended up being specifically rejected. It would be hypothesized that since the Western (or, more accurately, mass middle class) system of dating is now the default in both cultures, they will have similar strategies for relationship and mate selection.

Study 2

In study 2, the validity of the BSQ-K was examined with another set of Korean college student sample. Since breakup strategies involve individuals' efforts in a specific situation when they pursue termination of a romantic relationship, it can be related to their overall problem-solving or coping strategies. In this vein, Baxter (1982) maintained the existing association among the breakup strategies, conflict resolution strategies, and compliance-gaining strategies. For example, because rejectors in a relationship experience the reputational cost of being perceived by others being cruel (Perilloux & Buss, 2008), they are likely to necessitate effective problem-solving strategies to reduce stress. Based on this, this study examined the association between the breakup strategies and problem-focused coping strategies to establish the concurrent validity of the BSQ-K. Moreover, its relationship with social desirability (i.e., positive response bias) was examined to establish the discriminant validity of the scale. Social desirability is frequently used as an indicator of discriminant validity in scale construction (King & Bruner, 2000).

Method

Participants

A total of 174 Korean college students were recruited via online survey (52 men, 122 women, mean age of 21.59, SD = 2.48; range 18–43 years). Of the participants, the average age of men were 22.02 years (SD = 2.34; range 18–27 years), and women were 21.41 years (SD = 2.53; range 18–43 years). Eighty-three students were currently in a romantic relationship (47.7%), 91 students were single (52.3%), where 18 of 52 men (34.6%) were in a romantic relationship, and 34 students (65.4%) were single. For women, 65 out of 122 students (53.3%) were in a romantic relationship, and 57 students (46.7%) were single. The average number of romantic relationships of the participants was 2.90 (SD = 2.14; range 1–15). The average number of romantic relationships for men were 3.19 (SD = 2.56; range 1–15), and for women, 2.77 (SD = 1.93; range 1–13). The average duration of present or most recently terminated romantic relationship of the participants was reported at 12.29 months (SD = 11.43; range 0–60 months). The average duration of romantic relationship of men were 12.15 months (SD = 12.16; range 0–60 months), and for women, 12.34 months (SD = 11.15; range 0–48 months). Similar to Study 1, informed consents regarding research participation were obtained online.

Measures

The Korean Breakup Strategies Questionnaire (BSQ-K)

The 30-item BSQ-K finalized in Study 1 was utilized to measure breakup strategies. The internal consistency reliability coefficient (Cronbach's α) in Study 2 were Avoidance/withdrawal (.89), Indirect manipulation (.83), Open confrontation (.68), and Positive tone/self-blame (.69).

The Korean Version of the Problem-Focused Style of Coping (K-PF-SOC)

The PF-SOC, originally developed by Heppner, Cook, Wright, and Johnson (1995), validated by Lee and Park (2009) was utilized. The K-PF-SOC (18 items) consists of three factors (Reflective coping style, Suppressive coping style, and Reactive coping style). Reflective coping style (7 items) refers to planning and strategically coping with problem solving circumstances. Suppressive coping style (6 items) involves avoiding, and Reactive coping style (5 items) concerns exhausting one's energy while reacting emotionally and distorting problematic situations. Higher scores indicate frequent use of the specific coping style. The responses were rated on a 5-point Likert scale (1 = almost never to 5 = almost all of the time). The internal consistency reliability coefficients (Cronbach's α) of each sub-scale in Lee and Park's (2009) research were Reflective coping (.67), Suppressive coping (.75), and Reactive coping (.61). In this study, the reliability coefficients were Reflective coping (.77), Suppressive coping (.77), and Reactive coping (.64).

The Korean Short-Version of the Social Desirability Scale (K-SDS-9)

The SDS-17, originally developed by Stöber (2001), translated and validated by Bae, Lee, and Ham (2015) was utilized. The K-SDS-9 (total 9 items) was rated on a 5-point Likert- scale (1 = never to 5 = almost always). Higher scores indicate a higher possibility of responding in a socially desirable way. The internal consistency reliability coefficient (Cronbach's α) among items in Bae et al.'s (2015) study yielded .72, while in the current study it was reported as .60.

Analysis

An independent samples t-test was performed to examine if breakup strategies preference differs by gender. In addition, Pearson correlation analysis was conducted to explore the relationship among breakup strategies, three styles of coping (Reflective coping, Suppressive coping, and Reactive coping styles) and social desirability.

Results

Differences in breakup strategies preference by gender

In order to examine the gender differences in preferred breakup strategy, an independent samples t-test was implemented. The results showed significant differences by gender regarding Avoidance/withdrawal (t = −3.71, p < .001) and Open confrontation (t = −3.14, p = .002). That is, male participants were reported to use both Avoidance/withdrawal and Open confrontation comparatively more than female participants. By contrast, there was no significant gender difference regarding Indirect manipulation (t = .09, p = .927) and Positive tone/self-blame (t = 1.36, p = .175). Thus, Hypothesis 3 (the BSQ-K's sub-factors will differ by gender) was partially supported.

Correlations among breakup strategies, styles of coping, and social desirability

Table 2 presents results of the Pearson correlation analysis that was performed to examine whether the Breakup strategies show significant correlations with three styles of coping (Reflective coping, Suppressive coping, and Reactive coping styles) and social desirability. For Indirect manipulation, results showed a significant positive correlation with Suppressive coping (r = .21, p = .006). Also, Open confrontation showed a significant positive correlation with Reflective coping (r = .24, p = .002), whereas Avoidance/withdrawal and Positive tone/self-blame showed no significant correlation with styles of coping. Moreover, social desirability showed a significant positive correlation solely with the Positive tone/self-blame (r = .24, p = .002) of the Breakup strategy. Thus, Hypothesis 4 (there will be significant correlations between the breakup strategies and problem-focused coping strategies) and Hypothesis 5 (there will be no correlation between the breakup strategies and social desirability) were partially supported.

Table 2.

Correlation results among breakup strategies, problem-focused style of coping, and social desirability (N = 174).

	Avoidance/ withdrawal	Indirect manipulation	Open confrontation	Positive tone/ self-blame
Style of coping
Reflective	.08	.04	.24*	.13
Suppressive	.12	.21*	−.05	−.04
Reactive	.06	.08	.01	−07
Social desirability	−.02	−.12	.01	.24*

p < .05.

Discussion

The BSQ-K's validity was additionally examined with another group of Korean college students recruited through the online survey. Foremost, there were gender differences in breakup strategies preference measured by the BSQ-K. In particular, female participants preferred Avoidance/withdrawal and Open confrontation as compared to their male counterparts. Such results indicate that gender is a significant factor to understand the nature and pattern of Koreans' Breakup strategies use. In addition, examining the relationship between the BSQ-K and personal styles of coping provided an estimate of the concurrent validity of the BSQ-K. In conceptually expected ways, Indirect manipulation in breakup strategies use was associated with Suppressive coping, an indicator of avoidance in problem solving, and Open confrontation was related to Reflective coping that reflects, a planned and systematic problem-solving. Lastly, discriminant validity was established through a low correlation between the BSQ-K and a measure of socially desirable responding. The only exception was Positive tone/self-blame which showed a low correlation with social desirability, a conceptually expected result.

General Discussion

This study, consisting of two sub-studies, provides a culturally sensitive evaluation and measure of the BSQ in an Asian culture such as South Korea. The Korean version of the BSQ is sound in terms of several aspects of validity (construct, concurrent, and discriminant validity). These were some cross-cultural differences. For example, the merging of multiple factors in the original BSQ (Collins & Gillath, 2012) into one factor (Indirect manipulation) in South Korea reflects distinctive characteristics of Korean culture. In Western individualistic cultures, an individual is viewed as a separate entity. By contrast, collectivistic cultures such as South Korea consider the smallest unit of social construct as a relationship between individuals or groups (Cho, 2006) and emphasize interpersonal harmony and unity within inner group circle (Triandis, 1989, 1990). It is likely that this synthetic and holistic way of Korean thinking (Choi, 2008) was evinced in this study or maybe that the original four-factor structure was good anyway.

In the same context, characteristics of Korean culture that emphasizes “noon-chi” (Korean term which translates to social sensibility) could be related to this outcome. “Noon-chi” refers to the act and skill of inferring what others want through comprehensive evaluation of past encounters and the given situation of those individuals. It is a commonplace that Korean culture demands that its participants be very sensitive to reading others' needs. Individuals expect others to notice their needs in advance and act accordingly (Choi et al., 1999). Thus, Indirect manipulation factor is presumed to reflect the Korean characteristic of indirectly implying breakup, waiting for their partner to behave according to their expectations. “Dating” is a new concept in Western society, and it is definitely a byproduct of Western mass prosperity following the Industrial Revolution. It might be that the economic structure of “dating” and middle class for managing the hard parts of it, and middle class life simply imply a certain set of strategies for managing the hard parts of it, and individuals in each society implement them according to their own natures.

In addition, the results of the Rasch model analysis indicated that the 5-point Likert scale was more appropriate in all factors of the BSQ-K in comparison to the 7-point Likert-type scale used in the initial study. It may be useful to apply this more robust type of analysis to Western participants, to see if this result holds in general.

Limitations to the current study should be noted. First, although the factor structure of the BSQ-K was proven to be applicable for Korean college students, it is difficult to state that thorough validity and reliability estimates were established through associations with various criterion variables. Additional information on the result of using particular breakup strategies such as predictive validity, and test–retest reliability should be provided in future (possibly longitudinal) studies. Moreover, the sample of the current study did not cover diverse age groups or socioeconomic classes. Hence, there are clear limitations in generalizing the research results. Furthermore, the possible existence of an inherent breakup strategy reflecting Korean values and culture apart from the four factors of the BSQ-K cannot be excluded. Lastly, 32 students (11.0%) who have no breakup experience responded to the questionnaire hypothetically. For this reason, it is possible that their breakup strategy might differ in reality.

Despite such limitations, the current study provides meaningful implications, from a cross-cultural perspective, in expanding understanding of cultural universality and cultural specificity through testing the validity of the Western breakup strategy measurement tools with Korean college students. From a research perspective, the current study is significant in securing a tool to illuminate Korean college students' breakup strategies. This finding will serve as a foundation to open possibilities of future cross-cultural research on precursors or predicting variables of breakup strategies (e.g., personality, attachment, psychological schema, and gender) and outcome variables (e.g., psychological adjustment after breakup). Lastly, from a counseling practitioner perspective, the BSQ-K will provide guidance in objectively evaluating the college student clients' and their partners' breakup strategies, and designing effective intervention accordingly.

Footnotes

Article Notes

References

Anderson

E. B.

(1973) Conditional inference for multiple-choice questionnaires. British Journal of Mathematical and Statistical Psychology 26(1): 31–44.

Andrich

(1978) Rating formulation for ordered response categories. Psychometrika 43: 561–573.

Asparouhov

Muthén

(2009) Exploratory structural equation modeling. Structural Equation Modeling: A Multidisciplinary Journal 16(3): 397–438.

Banks

S. P.

Altendorf

D. M.

Greene

J. O.

Cody

M. J.

(1987) An examination of relationship disengagement: Perceptions, breakup strategies, and outcomes. The Western Journal of Speech Communication 51: 19–41.

Bae

B. H.

Lee

D.-G.

Ham

K. A.

(2015) Validation of the Korean short-version of social desirability scale (SDS-9): Using the Rasch model. Korea Journal of Counseling 16(6): 177–197.

Baxter

L. A.

(1982) Strategies for ending relationships: Two studies. The Western Journal of Speech Communication 46: 223–241.

Baxter

L. A.

(1984) Trajectories of relationship disengagement. Journal of Social & Personal Relationships 1: 29–48.

Baxter

L. A.

(1985) Accomplishing relationship disengagement. In: Duck

Perlman

(eds) Understanding personal relationships: An interdisciplinary approach, London, England: Sage, pp. 243–265.

Browne

M. W.

(1972) Oblique rotation to a partially specified target. British Journal of Mathematical and Statistical Psychology 25(2): 207–212.

10.

Browne

M. W.

(2001) An overview of analytic rotation in exploratory factor analysis. Multivariate Behavioral Research 36(1): 111–150.

11.

Byrne

B. M.

(2001) Structural equation modeling with AMOS: Basic concepts, applications, and programming, Mahwah, NJ: Lawrence Erlbaum Associates.

12.

Chang, H. J. (2013). Posttraumatic growth from romantic relationship breakup in college students: Mediating effect of intentional rumination between optimism and posttraumatic growth (Unpublished master's thesis). University of Chonnam, Chonnam, Korea.

13.

Cho

K. H.

(2006) East and west: Ideal types of human being, Seoul, Korea: Jisik-Sanup.

14.

Choi

K. H.

(2008) The Confucian origin of the East Asian collectivism, (pp. 150–151). Seoul, Korea: Jisik-Sanup.

15.

Choi

S. H.

Yoon

H. K.

Hahn

D. W.

Cho

K. H.

Lee

S. H.

(1999) Eastern psychology, Seoul, Korea: Jisik-Sanup, pp. 399–420.

16.

Cody

M. J.

(1982) A typology of disengagement strategies and an examination of the role intimacy, reactions to inequity and relational problems play in strategy selection. Communication Monographs 49: 148–170.

17.

Collins

T. J.

Gillath

(2012) Attachment, breakup strategies, and associated outcomes: The effects of security enhancement on the selection of breakup strategies. Journal of Research in Personality 46: 210–222.

18.

Comrey

A. L.

Lee

H. B.

(1992) A first course in factor analysis, (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates.

19.

Fox

C. M.

Jones

J. A.

(1998) Uses of Rasch modeling in counseling psychological research. Journal of Counseling Psychology 45(1): 30–45.

20.

Heppner

P. P.

Cook

S. W.

Wright

D. M.

Johnson

Jr. (1995) Progress in resolving problems: A problem-focused style of coping. Journal of Counseling Psychology 42: 279–293.

21.

Heppner

P. P.

Wampold

B. E.

Owen

Thompson

M. N.

Wang

K. T.

(2015) Research design in counseling, 4th ed. Belmont, CA: Cengage.

22.

Hong

(2000) The criteria for selecting appropriate fit indices in structural equation modeling and their rationales. Korean Journal of Clinical Psychology 19(1): 161–177.

23.

Hong

Kim

Wolfe

(2005) A psychometric revision of the European American Values Scale for Asian Americans using the Rasch Model. Measurement and Evaluation in Counseling and Development 37(4): 194–207.

24.

Hong

Wong

E. C.

(2005) Rasch rating scale modeling of the Korean version of the Beck Depression Inventory. Educational and Psychological Measurement 65(1): 124–139.

25.

Bentler

P. M.

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

26.

Kang, Y. H. (2016, 29 November). “Why won't you go out with me?” Woman in her twenties murdered by her colleague with a blunt instrument. Yonhapnews. Retrieved from the Yonhapnews website: http://www.yonhapnews.co.kr.

27.

King

M. F.

Bruner

G. C.

(2000) Social desirability bias: A neglected aspect of validity testing. Psychology and Marketing 17(2): 79–103.

28.

Kozan

M. K.

(1997) Culture and conflict management: A theoretical framework. International Journal of Conflict Management 8(4): 338–360.

29.

KT Economy & Management Institute. (2016). The mobile trend in the first half of 2016. Retrieved from the Yonhapnews website: http://www.digieco.co.kr.

30.

Lee

D.-G.

Park

H.-J.

(2009) Clusters of coping style and their differences in procrastination and psychological distress. Korea Journal of Social and Personality Psychology 23(2): 43–57.

31.

Lim, S. Y. (2016, 25 June). Abrupt breakups after enjoying kakaotalk dating. Joongang Daily. Retrieved from the Joongang website: http://www.joongang.co.kr.

32.

Linacre

J. M.

(2005) WINSTEPS Rasch measurement [computer software], Chicago, IL: Winsteps.com.

33.

Lorenzo-Seva

Frrando

P. J.

(2016) FACTOR 10.4. [computer software], Spain: University of Rovirai Virgili.

34.

LSinfo. (2015, 18 October). The reason why unmarried men and women struggle with anger management: A novel survey conducted by marriage agency Duo. Edaily. Retrieved from the Edaily website: http://www.edaily.co.kr.

35.

Marsh

H. W.

Muthén

Asparouhov

Lüdtke

Robitzsch

Morin

A. J.

Trautwein

(2009) Exploratory structural equation modeling, integrating CFA and EFA: Application to students' evaluations of university teaching. Structural Equation Modeling: A Multidisciplinary Journal 16(3): 439–476.

36.

McDonald

R. P.

(2005) Semiconfirmatory factor analysis: The example of anxiety and depression. Structural Equation Modeling 12(1): 163–172.

37.

Metts

Cupach

W. R.

Bejlovec

R. A.

(1989) I love you too much to ever start liking you': Redefining romantic relationships. Journal of Social and Personal Relationships 6(3): 259–274.

38.

Muthén

L. K.

Muthén

B. O.

(2010) Mplus (Version 6.0) [Computer software], Los Angeles, CA: Muthén & Muthén.

39.

Park

S. J.

(2015) The effect of Social support perceived by experienced interpersonal trauma on posttraumatic growth: Mediating effects of active stress coping. Journal of Future Oriented Youth Society 12(1): 25–47.

40.

Perilloux

Buss

D. M.

(2008) Breaking up romantic relationships: Costs experienced and coping strategies deployed. Evolutionary Psychology 6: 164–181.

41.

Rasch

(1960) Probabilistic models for some intelligence and attainment tests, Copenhagen, Denmark: Danish National Institute for Educational Research.

42.

Sbarra

D. A.

(2006) Predicting the onset of emotional recovery following nonmarital relationship dissolution: Survival analyses of sadness an anger. Personality and Social Psychology Bulletin 32: 298–312.

43.

Simon

R. W.

Barrett

A. E.

(2010) Nonmarital romantic relationships and mental health in early adulthood: Does the association differ for women and men? Journal of Health and Social Behavior 51(2): 168–182.

44.

Sperber

A. D.

Devellis

R. F.

Boehlecke

(1994) Cross-cultural translation: Methodology and validation. Journal of Cross-Cultural Psychology 25: 501–524.

45.

Sprecher

Zimmerman

Abrahams

E. M.

(2010) Choosing compassionate strategies to end a relationship: Effect of compassionate love for partner and the reason for the breakup. Journal of Social Psychology 41(2): 66–75.

46.

Stöber

(2001) The social desirability scale-17(SDS-17): Convergent validity, discriminant validity, and relationship with age. European Journal of Psychological Assessment 17(3): 222–232.

47.

Ting-Toomey

Yee-Jung

K. K.

Shapiro

R. B.

Garcia

Wright

T. J.

Oetzel

J. G.

(2000) Ethnic/cultural identity salience and conflict styles in four US ethnic groups. International Journal of Intercultural Relation 24(1): 47–81.

48.

Triandis

H. C.

(1989) The self and social behavior in differing cultural contexts. Psychological Review 96: 506–520.

49.

Triandis

H. C.

(1990) Cross-cultural studies of individualism and collectivism. In: Berman

(ed.) Nebraska symposium on motivation, 1989, Lincoln, NE: University of Nebraska Press, pp. 41–133.

50.

Velicer

W. F.

Eaton

C. A.

Fava

J. L.

(2000) Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In: Goffman

R. D.

Helmes

(eds) Problem and solutions in human assessment: Honoring Douglas N. Jackson at seventy, New York, NY: Springer Science & Business Media, pp. 41–71.

51.

Wright

B. D.

Masters

G. N.

(1982) Rating scale analysis, Chicago, IL: Measurement, Evaluation, Statistics, and Assessment Press.

52.

Wright

B. D.

Linacre

J. M.

Gustafson

J. E.

Martin-Löf

(1994) Reasonable mean-square fit values. Rasch Measurement Transactions 8(3): 370.