A Dimensional Approach to Characterizing On-Again / Off-Again Romantic Relationships

To better characterize on-again/off-again (on-off) relationships, the current study examined variations in these relationships using a dimensional approach. Based on a five-typology of on-off relationships, corresponding dimensions were developed and assessed in relation to breakup and renewal negotiations as well as relational quality characteristics. Canonical correlation analyses with 299 participants currently or previously in an on-off relationship revealed two major patterns: (1) positive on-off dynamics (e.g., relationship improvement, maintained interest) were associated with more successful transitions and greater relational quality, and (2) a combination of positive and negative on-off dynamics was related to positive change but also waning interest in the relationship and a greater difficulty negotiating relational transitions. Overall, the current findings suggest certain dimensions play stronger roles in characterizing on-off relationships, and assessing the dimensions in combination (i.e., those that covary in predicting relational characteristics) can reveal insights beyond those provided by categorical classifications of on-off relationships.

on the types are associated with relational characteristics to provide a more complex understanding of what qualities best characterize patterns in on-off relationships.This will, in turn, increase explanatory power in predicting the trajectories and dynamics of these relationships.
A Dimensional Approach to Characterizing On-Off Relationships Dailey et al. (2013) found five types of on-off relationships based on how partners negotiated or navigated the relational transitions (i.e., breakups and renewals).This lens was employed given that transitions are defining features of on-off relationships, they are critical points in on-off relationships, and the negotiation of these changes in relational status provides insights into the general nature of on-off relationships.To derive these types, the authors interviewed individuals who had been, or were currently, in an on-off relationship.They were asked to describe the communication surrounding each breakup and renewal in succession.In classifying the participants into the types, Dailey et al. qualitatively assessed the patterns that developed across the course of the relationship (e.g., whether the transitions resembled each other, whether the participant characterized both partners as approaching the transition similarly or differently).The types were further substantiated with a follow-up, quantitative analysis with a different sample.We base this dimensional approach on these five types as no other typologies or classifications exist regarding on-off relationships; this is the first exploration into the heterogeneity of on-off relationships.
The habitual type reflected relationships in which partners fell back into the relationship without much negotiation regarding the transitions.These partners renewed because it was relatively convenient and easy, and the on-off relationship provided companionship and comfort until one or both partners became interested in someone else.
The mismatched type encompassed relationships in which the cyclical nature was due to partners having incongruent internal qualities (i.e., personalities, desires) or to external factors (e.g., geographic distance, different schedules).For example, one partner would be interested in the relationship at one time and the other partner would be interested at a different time.Despite this, these partners tended to reported relatively better communication and relational dynamics than other types.For partners in the capitalized-on-transitions type, relational transitions were used to test the relationship, manage problems, or create opportunities to improve the relationship.
The on-off nature facilitated positive change in these relationships, and these partners reported the highest relational quality.Partners in the gradual separation type eventually realized the relationship was not going to work or they were no longer interested in continuing the relationship.As such, these partners reported having more closure about the ending of the relationship.In the controlling partner type, one partner tended to control the progression of the relationship or be more persistent in continuing the relationship, and the use of manipulation and control tactics was common.For example, one partner would degrade or confuse the other into believing they were at fault for the negative dynamics in the relationship.These partners reported lower relational quality as compared to the other types.
There is great heuristic and practical value in typologies of relationships (see, e.g., VanLear et al., 2006).They offer a parsimonious characterization of the inherent complexity of relationships, highlight differences and patterns in relationships, and yield more straightforward practical applications.Yet, the trade-off of parsimony is that a level of sophistication is lost when treating all relationships within a category or type as the same.Further, typologies tend to be better conceptual, rather than measurement, tools.An example of a typology that was transformed into dimensions is romantic attachment (cf., Brennan, Clark, & Shaver, 1998;Hazan & Shaver, 1987); researchers have moved from a three or four mutually-exclusive categorical measurement to two main dimensions of relational anxiety and avoidance.Thus, extending on Dailey et al.'s (2013) types, the current investigation characterizes on-off relationships using dimensions derived from the types.
Employing a dimensional approach to understanding on-off relationships is warranted for several reasons.First, whereas typologies often assume qualitative differences, dimensions assume quantitative differences (i.e., differences in degree; see Fraley & Waller, 1998).Rather than on-off relationships having wholly distinct patterns, our assumption is that on-off relationships vary in the degree to which they reflect or engender certain characteristics salient to this form of dating relationship.For example, all on-off relationships likely vary in how much partners work to improve the relationship, have power issues, or maintain interest in their partner.Accordingly, on-off relationships can be better characterized by their degrees of these dimensions rather than placing them in discrete categories.Second, in assessing the different on-off types, not all characteristics examined showed significant differences; further, when significant differences emerged, most showed the capitalized-on-transitions type was different from the others (see Dailey et al., 2013).Hence, the variations that exist may not emerge in analyses restricting relationships to one type.Yet, if the types are considered dimensions on which all on-off relationships vary, a more complete picture of on-off relationships may emerge.
Our overarching goal is thus to extend on Dailey et al. (2013) to substantiate patterns in on-off relationships by utilizing dimensions of each type rather than mutually exclusive categories.If there are different types of on-off relationships, the characteristics assessed should highlight these differences and reveal larger patterns to better understand the nature of variation within these relationships.For example, certain dimensions in combination may better reflect different types of on-off relationships (e.g., high capitalized-on-transitions combined with low habituation may be associated with higher relational quality, or high mismatched and control may be associated with lower relational quality) than mutually exclusive categories.Hence, we conceptualize each of the on-off types as a dimension, focusing on the key characteristics of each type, and explore how the dimensions are related to breakup and renewal characteristics as well as general relational quality factors through canonical correlation analyses.

Relational Transition Characteristics
Given that the on-off types were extracted based on the negotiation of relational transitions, we first assess how the on-off dimensions are associated with breakup and renewal characteristics.Relational transitions can occur in various ways (Mongeau & Teubner, 2002), and variability in how on-off partners manage their transitions should be reflected in the on-off dimensions.Based on Dailey, Rossetto, McCracken, Jin, and Green's (2012) conclusions regarding relational transitions in on-off relationships, the specific transition characteristics we assess, for both breakups and renewals, are: directness, explicitness, mutuality, struggle, and resolve.These characteristics should reflect differences in the on-off dimensions.
Directness is the degree of clarity regarding the enactment or initiation of a relational transition or the degree to which partners understand a breakup or renewal has occurred.Previous research shows both dissolution (e.g., Baxter, 1985) and reconciliation strategies (Bevan, Cameron, & Dillow, 2003) range from avoidance to explicit discussion.Explicitness is the degree to which partners overtly discuss the transitions and make the transition clear.Whereas directness is about the initiation of the transition, explicitness is about the negotiation and process of the transition.Mutuality is the level of agreement among partners regarding the transition.Most breakups in dating relationships are not mutual (e.g., Sprecher, 1994), particularly in on-off relationships (Dailey, Pfiester, et al., 2009); and partners are more persistent in reconciliation attempts when dissolutions were unilateral (Cupach & Metts, 2002).In addition, on-off partners often struggle with the relational transitions (Dailey, Rossetto, et al., 2012).Many couples maintain contact after breakups (Dailey, Rossetto, Pfiester, & Surra, 2009), but some partners report difficulty in determining the appropriate quantity and quality of contact after breakups.Finally, resolve pertains to partners' decisiveness in redefining the relationship, or in other words, their commitment in maintaining the most recent transition.To manage the transitions, some on-off partners allowed an "open door" after a breakup or declared trial renewals showing they viewed the transitions as malleable or temporary (Dailey, Rossetto, et al., 2012).
The on-off dimensions should be associated with these relational transition characteristics in certain ways.For example, Dailey et al. (2013) found the capitalized-on-transitions and mismatched types reported more directness as well as more mutuality regarding renewals, and the controlling type used less explicitness than other types.
Further, the capitalized-on-transitions type had more resolve about renewals than all other types.Additionally, certain partners should report greater struggle in managing the transitions (e.g., those who renew with a more controlling partner), whereas others should report less difficulty (e.g., breakups for those experiencing greater gradual separation).

Relational Quality Characteristics
Certain types appeared to involve greater relational quality (e.g., capitalized-on-transitions, mismatched) than others (e.g., controlling) (Dailey et al., 2013).Hence, we here assess how the on-off dimensions are associated with characteristics reflecting relational quality: relational uncertainty, quality of alternatives, feeling the relationship was special or unique, equity, and social network support.These characteristics have emerged as salient in onoff relationships and should theoretically vary with regard to the on-off dimensions.In other words, any differences in on-off relationships should be highlighted or revealed through these characteristics.
Relational uncertainty, or the degree of confidence in perceptions of involvement in relationships (Knobloch & Solomon, 2002), is one of the most prominent factors regarding on-off relationships.On-off partners are often uncertain about the status of the relationship after transitions, particularly breakups (Dailey, Jin, Pfiester, & Beck, 2011;Dailey, Rossetto, et al., 2009, 2012), and this uncertainty increases with greater renewals (Dailey, Pfiester, et al., 2009).Yet, greater uncertainty was linked with more commitment in on-off partners whose relationships were recently dissolved (i.e., in an "off" phase; Dailey et al., 2010).As such, relational uncertainty may be positively associated with certain on-off dimensions (e.g., habitual) but negatively associated with others (e.g., capitalizedon-transitions).
Generally speaking, when dating partners perceive better quality alternatives, they are likely to dissolve their current relationship (Attridge, Berscheid, & Simpson, 1995;Le, Dove, Agnew, Korn, & Mutso, 2010).A potential reason relationships become or remain on-off in nature is that partners do not perceive their alternatives as much higher in quality than their on-off partner (Dailey, Rossetto, et al., 2009;Dailey, Rossetto, et al., 2012).Some partners may wish to pursue their alternatives while maintaining their on-off relationship in some form.Yet, partners in on-off relationships have varying views of alternatives depending on their perceived stability (Dailey, Middleton, & Green, 2012).Thus, certain on-off dimensions may have positive associations with the quality of alternatives (e.g., gradual separation), whereas others may have negative associations (e.g., capitalized-on-transitions).
Feeling the relationship is unique or that the partner is "the one" is related to lingering feelings, which was the most predominant factor on-off partners reported as to why they renewed their relationship (Dailey et al., 2011).
Believing that a partner is "the one" is also similar to feelings of love which is associated with relational stability (see for reviews, Cate, Levin, & Richmond, 2002;Le et al., 2010) as well as difficulty in dealing with relationship dissolution (Sbarra & Emery, 2005).Further, of the characteristics assessed, Dailey et al. (2013) found the most differences among the types for lingering feelings-the capitalized-on-transitions and mismatched groups reported more lingering feelings than the gradual separation and controlling groups.As such, the capitalized-on-transitions and mismatched dimensions should be more strongly associated with feeling the relationship is unique.
Many romantic partners perceive an imbalance of inputs and rewards in their relationships (Hatfield, Utne, & Traupmann, 1979), and this inequity is linked with greater relational instability (Utne, Hatfield, Traupmann, & Greenberger, 1984).In on-off relationships, Dailey et al. (2011) surmised an imbalance of interest in some on-off couples given that one partner tended to initiate dissolutions and the other partner tended to initiate renewals.
Partners with more interest in pursuing or maintaining the relationship may perceive they are underbenefitted as compared to their partners (e.g., their partner is getting a better deal in the relationship), which often relates to feelings of depression (Sprecher, 2001a) and higher incidence of subsequent breakups (Sprecher, 2001b).Hence, (in)equity is likely an issue in many on-off relationships, perhaps particularly for those with a controlling nature, as well as associated with partners' overall sense of relational quality.
Given that social network support has been established as an important component of relational stability in romantic relationships generally (Agnew, Loving, & Drigotas, 2001;Sprecher & Felmlee, 2000), it is important to incorporate this into an analysis of how on-off relationships vary.For example, as the number of renewals increased in on-off relationships, network support decreased (Dailey, Pfiester, et al., 2009), and social network support has been associated with greater certainty about the relationship (Dailey, LeFebvre, Crook, & Brody, 2012).We purport that certain on-off dimensions will be positively associated with network support such as capitalized-on-transitions, whereas others will be negatively associated such as the habitual and controlling dimensions.
Overall, the purpose of this study is to assess variation or patterns in on-off relationships using a dimensional approach.Based on the types that emerged in Dailey et al. (2013), we examine how on-off dimensions are associated with transition negotiations (RQ1) as well as relational quality (RQ2).Explicating the variations in on-off relationships allows a more complex understanding of these relationships, which provides more explanatory power in predicting the ultimate outcomes of these relationships as well as greater precision in practical applications.

Participants and Procedure
Participants (n = 299) who were currently in or had previously experienced an on-off relationship (i.e., defined to participants as relationships that had broken up and renewed at least once) completed an online survey posted on the Amazon.comMechanical Turk website.Mechanical Turk is a website where workers complete tasks for nominal fees and that provides similar samples as other online recruitment tools or traditional recruitment methods (Buhrmester, Kwang, & Gosling, 2011).Participants were eligible for this study if they were currently in or had been in an on-off relationship, if their relationship began within the last five years (2006 at the time of this study), if they were at least 18 years of age, and if they resided in the U.S. Participants were paid $.50 for completing the survey.Our definition of on-off relationships (i.e., including those with only one renewal) allows a broader analysis to determine salient dimensions of these relationships.
Respondents ranged in age from 18 to 65 years old (M = 28.51,SD = 9.00).The sample consisted of 193 females and 106 males, and was predominantly Caucasian (72.9%), with the remainder of the sample being 10.4% Asian, 7.4% "Other", 5.4% African American, and 4.0% Hispanic/Latino.About half of the participants (n = 156, 52.2%) Interpersona  2013, Vol. 7(2), 196-214  doi:10.5964/ijpr.v7i2.131Dimensions of On-Off Relationships were still together, and about half (n = 143, 47.8%) were broken up.Number of renewals ranged from one to 20 times; when excluding the five outliers greater than 10 renewals, the number of renewals averaged 3.29 (SD = 1.90).The average length of relationships was 22.50 months (SD = 29.56).Number of renewals and relationship length did not significantly differ between those who were currently dating and those who were not, t s < 1.53, p s > .126.

Measures
On-Off Dimensions.A scale was created to assess on-off dimensions, one for each of the types found in Dailey et al. (2013): habitual, mismatched, capitalized-on-transitions, gradual separation, and controlling.Each author and an undergraduate research assistant individually created items for each type.Items were created to capture the essence of each type-isolating the different characteristic of each type.The authors, as a group, reviewed the items, discussing how central and reflective each item was to the type.The items were compiled and similar or identical items were combined.A final set of 60 items was chosen.Based on a pilot test with 226 participants, i 32 items that best captured the key characteristics for each type were selected for the main analyses (see Table 1).
The current sample rated their agreement with the selected 32 items on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).A confirmatory factor analysis (CFA) with the items loading on their representative factors yielded less than satisfactory fit, χ 2 (454) = 994.57,p < .001,CFI = .87,RMSEA = .06.We conducted an EFA which extracted six factors explaining 67.29% of the variance.The capitalized-on-transitions and gradual separation types clearly emerged.Two factors emerged for the mismatched type: incongruencies based on external factors (e.g., geography, network influences) and incongruencies based on internal factors (e.g., personality or relationship issues), which makes conceptual sense given that two subtypes were found by Dailey et al. (2013).However, the habitual items did not load highly on any of the factors and appeared to be obscuring the other factors.As such, we conducted a subsequent EFA excluding the habitual items.Five factors emerged that explained 57.90% of the variance (eigenvalues = 7.89, 3.65, 1.67, 1.28, and 1.14).Each of the remaining types clearly emerged (see Table 1).The items were thus combined into five on-off dimensions: mismatched -internal, mismatched -external, capitalized-on-transitions, gradual separation, and contention.All items were included in each factor with the exception of two control items that loaded on the mismatched external factor.With the exclusion of these two items, the control dimension reflects a contentious or manipulative, rather than a controlling, dynamic and the name was changed as such.Other than the two items excluded from the control factor, no other cross-loadings exceeded .30.It was clear when we renewed (e.g., I had no uncertainties about whether we were back together). .59 The breakup(s) helped us make changes in ourselves or in the relationship. .73 My partner and I set rules or guidelines about our relationship before we got back together. .76 The breakup(s) helped me or my partner realize that we wanted this relationship. .64 We explicitly discussed the nature of our relationship when we renewed. .80 Our breakup(s) were stressful, but usually ended up improving our relationship. .76 We worked to make our relationship stronger. .75 When we renewed, I knew our relationship would be stronger. .78 Mismatched --External (M = 4.29, SD = 1.23, α = .48) Factors such as living in different cities, school, or work were a major reason for the on-off nature in our relationship. .67 We would be happy together (would have been happy) if some external conditions changed.
. One or both of us used manipulation in the relationship .52 We often showed negative emotions even when we talked about getting back together. .58 I would characterize our relationship as "love-hate." .74 We sometimes failed to break up because one of us insisted on staying together. .69 The power in our relationship seemed to be imbalanced.† .36 .54-Sometimes I felt like I had no control over the direction of our relationship.† .

Habitual ‡
Whenever we renewed, I had my doubts about whether the relationship would work.
We tended to renew out of habit.
Our issues never seemed to be resolved.I stayed with my partner because s/he was better than any alternatives I had.
It was sometimes unclear whether we had broken up or not.
Whenever we renewed, I had my doubts about whether the relationship would work.Note.†Excluded from the calculation of the Contention score.‡Habitual items were excluded from the final exploratory factor analysis.2011) on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree).The items included: "We have a special connection", "I knew my partner was 'the one'," and "There was something special about this relationship".Reliability was high (α = .84;M = 4.94, SD = 1.59).Quality of alternatives was assessed with five items (e.g., "The people other than my partner with whom I might have become involved were very appealing") from Rusbult, Martz, and Agnew (1998) on a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree; α = .87;M = 4.02, SD = 1.52).We operationalized power through its conceptual inverse-equity.Relational equity was measured with two Likert-type items from Sprecher (1986) assessing the degree to which one partner is getting a better deal (i.e., 1 = "I have had a much better deal than my partner", 7 = "My partner has had a much better deal than I") and who is contributing more to the relationship (i.e., 1 = "I was much more likely to be the one to contribute more", 7 = "My partner was much more likely to be the one to contribute more").These two items were moderately correlated (r = .56;M = 2.50, SD = 0.98) and we recoded and combined them to create an overall score.Higher scores indicate greater equity in the relationship.
Transition Negotiation.Based on Dailey, Rossetto, et al. (2012) and Dailey et al. (2013), five breakup and renewal characteristics salient to on-off relationships were assessed.Similar to the on-off dimensions, the authors created several items for each dimension and selected the items that best captured the transition characteristics.

Results
The research questions pertain to relationships between sets of variables.As such, we employed canonical correlations.Canonical correlation analysis (CCA) is a multivariate analysis that extracts a synthetic (i.e., unobservable, latent) variable from each set of variables to maximize the bivariate correlation between the synthetic variables.
The advantages of CCA are that it decreases Type I error and can capture the complexity of the inter-relationships among the variables in a more parsimonious way than separate tests for each dependent variable (Sherry & Henson, 2005), which is a major goal of this study.Although we model the on-off dimensions as the independent variables (synthetic predictor) and the transition characteristics and relational quality variables as the dependent variables (synthetic criterion), this analysis is based on correlation and causality cannot be determined.
CCA yields as many canonical correlations or functions as the number of variables in the smallest set of variables (i.e., five in the current analyses).The first canonical function maximizes the association between the two synthetic variables, and subsequent functions maximize the correlation between additional synthetic variables from the variance not explained by previous functions.As recommended (Sherry & Henson, 2005), we assessed both the standardized canonical function coefficients (i.e., how observed variables load on the synthetic variable controlling for the other variables, similar to β in regression) and the structure coefficients (i.e., correlations between the observed and synthetic variables, similar to Pearson r) in determining the variables that contributed most to the synthetic variables.We present both coefficients in the figures.
iii RQ1 concerns the relationship between on-off dimensions and transition negotiation characteristics.Breakup and renewal characteristics were assessed in two separate canonical correlation analyses.In the breakup model, the full model across all functions was statistically significant, Wilks's λ = .504,F( 25 A graphical representation of the first two functions is presented in Figure 1.For the first function, the capitalizedon-transitions, gradual separation, and contention dimensions were more salient in the synthetic predictor; capitalized-on-transitions was negatively associated with gradual separation and contention.Explicitness most strongly contributed to the synthetic dependent variable, followed by directness, struggle, and resolve; lower explicitness and directness was related to higher resolve and struggle.In more general terms, this first function shows that the strongest pattern among the on-off dimensions and breakup characteristics was: partners who felt the relationship had not improved, became less interested in the relationship over time, and had a more contentious relationship (labeled downward spiral in the figure) also reported having breakups that were less explicit and clear, and with greater struggle and resolve (i.e., more ambiguous and difficult breakups; labeled avoidant but resolute breakups).
The second function showed the gradual separation, contention, capitalized-on-transitions, and internal mismatch dimensions relatively strongly, and positively, contributed to the synthetic variable.In addition, all breakup transition variables except directness were rather strongly related to the synthetic criterion variable; higher resolve, explicitness, struggle, and mutuality were positively related to one another, and resolve contributed most to the synthetic variable.This function overall suggests, with the remaining variance not explained by the first function, on-off relationships characterized by more manipulation, partners losing interest, and perceived incongruencies (albeit with positive change) were related to breakups that were more explicit, resolute, and mutual, but with more struggle as well.In other words, the second function reflects a second pattern of on-off relationships regarding the relationship between on-off dimensions and breakup characteristics: those whose interest in the relationship declined despite experiencing positive change (i.e., lost interest despite improvements) reported clearer, but also more trying, breakups (i.e., explicit but difficult breakups).In the renewal model, the full model across all functions was statistically significant, Wilks's λ = .266,F(25, 1075.09)= 18.42, p < .001,and explained 73% of the variance shared between the two variables sets.Notably, the renewal transition characteristics explained more variance than the breakup characteristics (i.e., 73% compared to 50%).
We report the results of the first two functions (Figure 2) as these were more meaningful based on the amount of variance explained-64.5% by the first and 6.2% by the second.
For the first function, the coefficients indicate that capitalized-on-transitions, gradual separation, and internal mismatch primarily contributed to the synthetic predictor; capitalizing was negatively related to gradual separation and internal mismatch.All five transition variables strongly contributed to the synthetic dependent variable, but explicitness, resolve, and struggle were more salient after controlling for one another (i.e., based on the standardized coefficients); higher explicitness and resolve were related to lower struggle in renewals.In sum, those who felt the relationship had improved and had similar goals for the relationship as their partners (i.e., positive progression) were also able to more explicitly negotiate renewals with less reservations (i.e., successfully negotiated renewals).
The second function showed internal mismatch, capitalized-on-transitions, gradual separation, and contention positively contributed to the synthetic predictor, and struggle and explicitness were the strongest components of the renewal characteristics synthetic criterion variable.These results suggest that those who were decreasingly interested in the relationship and experienced more manipulation, but were on the same page about the relationship with their partners (i.e., work did not overcome negatives) tended to have more explicit and difficult renewal interactions (i.e., labored renewals).RQ2 addresses the relationship between the on-off dimensions and relational quality.A similar canonical correlation analysis as those for RQ1 was conducted with the five relational quality variables entered as dependent variables.
The full model was statistically significant, Wilks's λ = .279,F(25, 1075.09)= 17.62, p < .001,and explained 72% of the variance shared between the two variables sets.Similar to the analyses above, the first two functions are interpreted; the variance explained by the two functions was 62.4% and 6.3%, respectively.
For the first function (see Figure 3), the coefficients indicate all of the on-off dimensions except external mismatch substantially contributed to the synthetic predictor.The capitalized-on-transitions dimension was negatively related to the gradual separation, contention, and internal mismatch dimensions.The more relevant dependent variables were relationship uniqueness, relational uncertainty, network support, and equity; less relational uncertainty and more network support were related to feeling the relationship was unique and equitable.Collectively, those who had maintained interest for the relationship and experienced positive change and less contention (yet also feeling they and their partner had similar goals for the relationship) also reported higher relational quality.Hence, those reporting positive progression also reported higher quality relationships.
The second function showed external and internal mismatch, capitalized-on-transitions, and gradual separation contributed most to the synthetic predictor; being internally and externally mismatched were negatively related to the other two variables.Relational uncertainty, relationship uniqueness, quality of alternatives, and network support most contributed to the synthetic dependent variable, with alternatives and support negatively related to the other two variables.Overall, this function suggests on-off partners who had a continued desire for the relationship despite not experiencing improvements or the partners having similar goals (i.e., maintained interest despite no change) appeared to feel the relationship was unique with fewer alternatives but had more uncertainty and less network support (i.e., unique but uncertainty and no support).

Discussion
Research on on-off relationships has progressed in its conceptualization of on-off relationships from a parsimonious, yet relatively uniform view of these relationships, to a more delineated approach by establishing an on-off typology (e.g., Dailey et al., 2013).Relational typologies are advantageous as they allow for a refined understanding of on-off relationships; however, mutually exclusive categories may obscure important complexities and patterns in the relationships.The focus of this research is on characterizing on-off relationships using a dimensional approach which illuminates variations in these relationships that may be less apparent when relationships are restricted to one category or type.

On-Off Relationship Dynamics: Two Major Patterns
As might be expected, on-off dimensions that denoted greater efforts and interest in the relationship were associated with more clearly negotiated transitions and greater relational quality.In the first (primary) functions, if we reverse the signs for the breakup CCA to facilitate interpretation and for consistency across the analyses (i.e., so that more positive on-off dimensions are associated with more positive transition and relationship characteristics), the findings suggest on-off partners who work to improve the relationship, maintain interest in the relationship, and minimize contention feel more hopeful about creating a more stable relationship.In particular, they negotiate their transitions more openly and with less difficulty, and their renewals indicate little hesitation and more commitment to continuing the relationship.Further, these partners have more certainty and confidence in the relationship as well as approval from their social network.Couching these findings within the previous on-off research provides greater insight on how or which on-off relationships transform to a state of greater stability.Although on-off relationships tend to have lower relational quality as compared to relationships that do not break up and renew (Dailey, Pfiester, et al., 2009), several studies suggest some couples believe the on-off nature of their relationship is beneficial because the transitions allow an opportunity to change the relationship or start fresh (Dailey et al., 2013;Dailey, Rossetto, et al., 2009, 2012).Separately, lingering feelings or a continued attachment to the partner has also been a predominant reason for renewals (Dailey et al., 2011).But the current results imply that the combination of working on the relationship and an increased, or at least maintained, interest in the relationship, perhaps along with minimizing manipulation, may be especially important.These partners may have more positive dynamics entering transitions as well as an enduring attachment to the partner which may facilitate positive change in the relationship.These are likely the couples who set terms and conditions during their renewals and/or report positive change during the transition process (see Dailey, Rossetto, et al., 2012), and as such, may be more likely to achieve relative stability (see also, Dailey, Middleton, et al., 2012).Conversely, those with less change and decreasing interest may be on the path towards permanent dissolution with their lower relational quality, avoidance of discussions about the transitions, and greater relational uncertainty.This aligns with research that shows couples who are uncertain about their relationship avoid conversations about their relationship (Afifi & Burgoon, 1998;Baxter & Wilmot, 1985), and avoidance, in turn, is related to lower relational satisfaction (Caughlin & Golish, 2002;Roloff & Ifert, 2000).
A second pattern emerged in all three of the second functions wherein partners reported a mix of positive and negative on-off dynamics.This pattern should be given less weight as the second functions explained relatively less variance than the first functions.Yet, given that this trend emerged in all three CCAs, it warrants further consideration and investigation.Specifically, despite working on the relationship and perhaps also feeling compatible with the partner, their interest in the relationship waned.This is reflected in greater commitment to breakups, less relationship uniqueness, and better alternatives.Yet, these partners appeared more conflicted about their relationships as evidenced by their greater struggle with both breakups and renewals.Perhaps this struggle occurs because, objectively-speaking, the relationship should be more satisfying (and their social network is in favor of the relationship as well).In social exchange terms (Thibaut & Kelley, 1959), the relationship perhaps never meets expectations or standards.This may be a reason for the on-off nature of the relationship-partners see the potential and thus make several attempts at making the relationship work but are never fully satisfied.The converse of this pattern is a maintained interest despite not experiencing any improvements or feeling mismatched with the partner.This is also related to perceiving fewer alternatives which may be driving the continued attachment to the relationship despite the negatives (see, e.g., Thibaut & Kelley, 1959).Further, these partners may have more difficulties in contracting their identities or redefining their self-concepts when separating from their partners (Lewandowski, Aron, Bassis, & Kunak, 2006;Slotter, Gardner, & Finkel, 2010), which draws them back to the relationship.These couples may be those who continue cycling as opposed to achieving permanent stability or dissolution (Dailey, Middleton, et al., 2012).These partners may see their relationship as sufficient but not fully meeting their standards.They also reported their alternatives as more attractive.Hence, as discussed within a social exchange framework by previous on-off research (Dailey, Middleton, et al., 2012;Dailey, Rossetto, et al., 2009), the current alternatives may not be so attractive as to warrant terminating the relationship, and partners may be waiting for a more attractive partner to materialize to fully dissolve the on-off relationship.Previous research suggests they may accomplish this by circumventing an explicit or permanent transition with strategies such as leaving the door open after breakups and viewing renewals as temporary or a trial (see Dailey, Rossetto, et al., 2012).In sum,

Theoretical and Practical Implications
The two dimensions that emerged as particularly important were the capitalized-on-transitions and gradual separation dimensions; both loaded on all synthetic variables in the first as well as second functions.Further, it was the combination of these two factors that appears to reflect the nature of on-off relationships.In the first functions, positive change was combined with less gradual separation (i.e., maintained or increased interest).Conversely, positive change was combined with decreasing interest in the second functions.In addition, the contention dimension significantly loaded on several synthetic variables.Interestingly, the habitual dimension did not emerge as a separate factor in our measure.Perhaps habituation is captured in other dimensions or a habitual nature is endemic to most on-off relationships.Additionally, although our initial conceptualization of the mismatched dimension was divided into internal (e.g., incompatible personality, desires) and external (e.g., geographic distance) aspects of the relationship, these dimensions did not play much of a role in the patterns found in the current data.Thus, based on this initial exploration of dimensions, the nature of on-off relationships may be best, and more parsimoniously, captured by the capitalized-on-transitions and gradual separation dimensions.Ultimately, we hope further research aids in solidifying the key factors that differentiate and characterize on-off relationships.
For those involved in on-off relationships, this research suggests several applications.For example, research on on-off relationships suggests some partners use breakups to manage problems (Dailey et al., 2013) and ultimately develop relational growth through conflict (see also Braiker & Kelley, 1979).If we assume that using relational transitions to test, manage, or create opportunities to improve the relationship (i.e., capitalized-on-transitions) represents the most positive and functional on-off dimension, then relational partners may be able to capitalize on the transitions through explicit negotiation; breakup explicitness and directness were related to less resolve and struggle, and renewal explicitness was related to greater resolve and less struggle.In general then, greater openness and clarity during the transitions may facilitate less difficult transitions as well as help stabilize the relationship.
Yet, if partners avoid or are equivocal in negotiating the transitions, uncertainty likely arises or increases.This uncertainty may intensify partners' indecisiveness regarding the relationship or foster detrimental communication patterns given that research on the relational turbulence model has found uncertainty is linked with more topic avoidance, indirectness, and negative appraisals (Knobloch & Carpenter-Theune, 2004;Knobloch & Solomon, 2005;Theiss & Solomon, 2006).Additionally, relational uncertainty may be associated with biasedly attributing the instability to the partner's personality or actions as opposed to their own, which may perpetuate maladaptive patterns or hinder opportunities to enact positive change (e.g., distress-maintaining vs. relationship-enhancing attributions; see Fincham, Harold, & Gano-Phillips, 2000).Ambiguity may also prolong the on-off relationship (Dailey, Rossetto, et al., 2009) when permanent dissolution would be best for one or both partners.Overall, clarity and openness may foster more successful transitions and greater change in the relationship.

Limitations and Future Research Directions
In the current study, we assessed one partner's perceptions of the relationship to capture on-off partners' perceptions of their on-off dynamics as well as transition and relational characteristics.Although one partner's perceptions of the relationship may be better for accessing certain relational qualities, obtaining data from both partners would be beneficial, particularly when focusing on communication dynamics (e.g., negotiation of transitions).In addition, the time since the beginning of the relationship (i.e., up to five years prior to participation) may have affected the recall of participants' experiences, particularly if they had terminated the relationship and began a relationship with a new partner.Further, this analysis was based on correlation, and thus, causality between the dimensions and transition and relational qualities cannot be determined.Research utilizing a longitudinal design would better capture causal relationships in this process.
An important contribution of this study is the scale assessing dimensions of on-off relationships.Although the items were reliable for most of the on-off types, additional research would help further refine and validate this measure.In addition, the original on-off types are based on partners' pattern of negotiating the transitions across the relationship; typologies based on other dynamics of on-off relationships may reveal other dimensions.Further, assessing the dimensions in relation to characteristics other than transitions or relational quality may yield other patterns.Measures for transition negotiation dimensions (e.g., explicitness, mutuality) were also created for the purposes of this study, and some of the dimensions yielded less than desirable reliabilities.Future research is also needed to refine the measures of these variables.

Conclusion
The current study offers several contributions.First, we assessed the types using a quantitative measure allowing partners' relationships to vary along these features of on-off relationships.This also allowed the simultaneous assessment of the dimensions revealing that combinations of dimensions covaried to predict relational characteristics, thus pointing towards more general dimensions or patterns that will better characterize different types of on-off relationships.From this, two major patterns emerged: 1) couples entering transitions with a stronger relational foundation are likely to weather the transitions better; these partners navigate their transitions with more openness and explicitness and appear to capitalize on the opportunities transitions offer in transforming the relationship; and 2) other on-off partners appear to experience a mix of positive change but waning interest and have more difficulty in navigating the transitions.In addition, the analyses showed certain dimensions were not as prominent (i.e., habitual, mismatched) suggesting certain features of on-off relationships may be relatively applicable to all on-off relationships or are less salient in differentiating types of on-off relationships.Overall, the current findings provide an additional step towards identifying factors or dimensions that best differentiate on-off relationships.Ultimately, isolating types or patterns in on-off relationships will enable better predictions and explanations of which on-off relationships will achieve stability and which will permanently dissolve.

Notes
i) The pilot sample was 226 undergraduate students from a large, Southwestern university.The sample was two-thirds female (68.6%) and the average age was 20.32 (SD = 2.14, range was 18-46).A little less than half of the sample was currently dating their on-off partner (43.8%).Participants reported their relationships had an average of 2.56 renewals (SD = 1.42) and lasted 21.88 months on average (SD = 15.26).An initial exploratory factor analysis (EFA) with an oblimin rotation yielded 14 factors with eigenvalues greater than 1.0.A subsequent EFA constraining the factors to five explained 46.40% of the variance (eigenvalues ranged from 2.16 and 14.45).Although not all items loaded on their respective factors, each type was represented.ii) All items measuring the transition negotiation characteristics can be obtained from the first author.iii) Additional output regarding the canonical correlations (e.g., structure coefficient, squared structure coefficient) can be obtained from the first author.The analysis provides more statistics helpful in interpreting the results: the canonical correlation coefficients (R c ) show the bivariate association between the two synthetic variables (similar to a Pearson correlation with the square of this being the amount of variance shared by the two synthetic variables), and squared structure coefficients (i.e., variance shared between the observed variable and the synthetic variable; the effect size, r 2 ).In addition, although relational status (i.e., together or broken up) is often related to participants' report of their relational qualities, including relational status in the criterion synthetic variable in the CCAs did not meaningfully alter the results in any of the analyses.Accounting for the number of renewals (i.e., including renewals as a criterion variable) or marital status also did not substantially alter the results.As such, in order to maximize power, we present the results with the full sample.
62 I'd like to (I wanted to) be with my partner, but circumstances keep getting in the way..78Mismatched --Internal (M = 4.98, SD = 1.23, α = .71)My partner and I tend(ed) to be interested in the relationship at different times..62 -We sometimes weren't "on the same page" in our relationship..73-At certain points, the timing just wasn't right in our relationship..56-At times, we wanted different things out of the relationship..72 -Contention (M = 4.31, SD = 1.28, α = .75) Quality.Network support was measured with Sprecher and Felmlee's (2000) six items addressing the extent to which friends and family members supported the relationship (e.g., "To what degree do you think your family disapproved/approved of the relationship"; 1 = very much disapproves, 7 = very much approves).Reliability was high (α = .86;M = 4.14, SD = 1.44).The extent to which the participant considered the relationship Interpersona 2013, Vol.7(2), 196-214 doi:10.5964/ijpr.v7i2.131Dimensions of On-Off Relationships special or unique was measured with three items stemming fromDailey et al. ( , 1075.09) = 8.71, p < .001,and the overall effect size (R c 2 ) was .496(1 -λ), showing that the full model explained 50% of the variance shared between the two variable sets.Although three models out of five were statistically significant, we regard only the first two functions as meaningful considering the variance explained and the canonical correlations among the synthetic variables.The squared canonical correlations (R c 2 ) of the first function indicate that it explained 35.7% of the variance, and the second function explained an additional 7.6% of the original observed variance.

Figure 1 .
Figure 1.First and second functions for breakup negotiation characteristics.Note.Standardized canonical function coefficients (structure coefficients) shown.The dashed arrows are for the second function.Relatively stronger coefficients are bolded (based on both standardized and structure coefficients).

Figure 2 .
Figure 2. First and second functions for renewal negotiation characteristics.Note.Standardized canonical function coefficients (structure coefficients) shown.The dashed arrows are for the second function.Relatively stronger coefficients are bolded (based on both standardized and structure coefficients).

Figure 3 .
Figure 3. First and second functions for relational quality.Note.Standardized canonical function coefficients (structure coefficients) shown.The dashed arrows are for the second function.Relatively stronger coefficients are bolded (based on both standardized and structure coefficients).

Table 1
On-Off Dimension Items and Factor Loadings