Delineating the Boundaries Between Nonmonogamy and Infidelity: Bringing Consent Back Into Definitions of Consensual Nonmonogamy With Latent Profile Analysis

Abstract

To gain insight into the mixed findings surrounding Consensual Nonmonogamy (CNM), this study developed the Triple-C model of commitment, conceptualizing relationship structures with three key dimensions: mutual consent, communication, and comfort. Latent profile analyses in an online sample (N = 1,658) identified five classes of relationship structures: two monogamous groups (68%; representing earlier- and later-stage relationships), CNM relationships (7.7%, marked by low interest in monogamy and high levels of mutual consent, comfort, and communication around commitment and EDSA), partially-open relationships (13%, with more mixed attitudes toward monogamy and lower consent, comfort, and communication), and one-sided EDSA relationships (11%, in which one partner desires monogamy while the other partner engages in EDSA with low levels of mutual consent, comfort, and communication). The monogamous and CNM groups demonstrated high levels of relationship and individual functioning, whereas the partially-open and one-sided nonmonogamous groups demonstrated lower functioning. These findings highlight the diversity of nonmonogamy that likely exists within self-report classifications like “swingers” and “open relationships,” providing a possible explanation for the mixed findings in previous work. Decision tree analyses identified a 4-item algorithm (COMMIT4) that classifies individuals into these groups with 93% accuracy, offering a tool for incorporating relationship structure diversity in future work.

We use the term nonmonogamy as an umbrella term to refer to a diverse phenomenon (e.g., open relationships, swinging, polyamory) involving both sexual and emotional aspects. The current study focused on sexual nonmonogamy in an effort to characterize an underlying facet of this phenomenon that is common to most forms of nonmonogamy. Although as many as 20% of individuals engage in open sexual relationships at some point in their lives (Haupert, Gesselman, Moors, Fisher, & Garcia, 2016), the reality of maintaining a healthy relationship with a nonmonogamous relationship structure can be challenging (Denfeld, 1974). Such couples must navigate the introduction of new sexual partners into their relationships in the context of a monogamy-focused culture, and would need to protect one another from possible feelings of jealousy and the judgment of others (e.g., Denfeld, 1974; De Visser & Mcdonald, 2007). Although previous work in this area (see Barker & Langdridge, 2010; Moors, Matsick, & Schechinger, 2017; Rubel & Bogaert, 2015 for reviews) has provided a foundation for understanding how non-traditional relationship structures can potentially benefit and harm romantic relationships, it has been limited by discrepancies in how these relationship structures are assessed and categorized across studies. Often relationship structures (i.e., monogamous vs. nonmonogamous) are assessed with just one or two items. Although this can simplify the classification of participants into relationship groups, such strategies limit researchers’ abilities to assess key aspects of relationship structures, including mutual consent, comfort, and open communication. In the current study, we therefore sought to examine these aspects of relationship structures more thoroughly, identifying fundamental relationship types and thereby delineating when Extra Dyadic Sexual Activity (EDSA – meaning any sexual activity with a person other than one’s primary romantic partner) can represent a healthy expression of sexuality in romantic relationships.

Previous Conceptualizations of Nonmonogamy

The Broader Phenomenon of Nonmonogamy

We use the term “nonmonogamy” to refer to the wide range of possible non-traditional relationship structures, reserving the term consensual nonmonogamy (CNM) for open relationships with high levels of mutual consent, mutual comfort, and ongoing communication surrounding EDSA. Within this conceptual framework, swinging (e.g., Bergstrand & Williams, 2000; Platteau, van Lankveld, Ooms, & Florence, 2017), polyamory (e.g., Conley, Matsick, Moors, & Ziegler, 2017; Mitchell, Bartholomew, & Cobb, 2014), and open relationships (e.g., Blasband & Peplau, 1985; Kurdek & Schmitt, 1986; Rubin & Adams, 1986) represent specific forms of nonmonogamy with different agreed upon contexts for EDSA. As we wanted to develop a broader conceptual understanding of the phenomenon of nonmonogamy, we chose to broaden the assessment of nonmonogamy in a manner that would encompass all forms.

Forms of Nonmonogamy

Previous work on nonmonogamy has primarily focused on three forms of nonmonogamy: open relationships, swinging, and polyamory. Open relationships typically refer to couples in which the partners have agreed to engage in EDSA while maintaining their dyadic relationship as their primary emotional bond (e.g., Blasband & Peplau, 1985). Couples in open relationships generally make specific arrangements about the acceptable context for EDSA (e.g., only bringing in third partners when together, avoiding mutual friends or ex-partners), and can still consider EDSA a betrayal if those conditions are violated (Hosking, 2013). Swinging represents another form of nonmonogamy in which romantic dyads agree to engage in EDSA in the context of social events (e.g., parties, destination vacations) focused on engaging in sexual activity with other similarly minded couples. Swingers also establish ground rules for when EDSA is acceptable (e.g., complete partner swapping, engaging only in 4-somes with other couples) and what activities are allowable (e.g., Bergstrand & Williams, 2000). Finally, polyamory has been defined as partners who are typically in emotionally and/or sexually intimate relationships with more than one person (e.g., Wood, Desmarais, Burleigh, & Milhausen, 2018). These three forms of nonmonogamy therefore differ (1) in how couples self-identify and (2) in the specific contexts that EDSA is allowable. However, they can occasionally also differ (3) in the presence of a primary dyad and the degree of emotional commitment outside of that dyad, and (4) in the level of communication surrounding the relationship structure and EDSA to all sexual and romantic partners (Balzarini et al., 2017). Although previous studies have often focused on just one of these forms of nonmonogamy, they have also often collapsed across categories to create a broader CNM group. To the degree that these different forms of nonmonogamy might present unique benefits or challenges to relationships, collapsing across all forms could obscure meaningful differences.

A New Conceptual Framework: The Triple-C Model

We sought to build on that previous work by augmenting the assessment of both monogamous and nonmonogamous relationship structures with processes we thought would be critical to the health and vitality of those relationships. To identify key processes, we drew upon the vulnerability-stress-adaptation (VSA) model of relationship functioning (Karney & Bradbury, 1995), which posits that (1) enduring vulnerabilities couples bring into romantic relationships, (2) stressful life events that couples face in the course of their relationships, and (3) adaptive and maladaptive methods of interacting with one another (e.g., conflict, support, forgiveness) are processes that interact with one another to ultimately promote sustained relationship health or gradual relationship decay. From this framework, we specifically focused on three fundamental processes that form the essential building blocks of healthy and vital relationship structures. Thus, we conceptualized both a lack of mutual (1) consent and (2) comfort with the current relationship structure as key enduring vulnerabilities that would likely generate ongoing stressful events for couples and trigger ongoing conflict and maladaptive relationship processes, ultimately resulting in relationship decay. We also conceptualized (3) mutual, ongoing, constructive and open communication surrounding the relationship structure as a key adaptive process that would help to buffer relationships from the adverse effects of enduring vulnerabilities and stressful events across time. Extending this further, EDSA that occurs in violation of the current agreed-upon relationship structure (or more broadly within relationship structures in which levels of the Triple-C components are low) would represent stressful events that would trigger the types of increasingly maladaptive cycles of aversive conflict that have been linked to relationship decay via both the VSA framework and by social learning theory (Koerner & Jacobson, 1994). We believed that these critical building blocks would offer a conceptual framework that would help characterize adaptive and maladaptive relationship structures within any self-identified group – be that monogamous couples, swingers, polyamorous relationships, or open relationships. It is worth noting that our analyses were fundamentally exploratory in nature and so we could not be sure of the exact nature of the groups that would emerge. However, these processes and the Triple-C model provided a foundation for that exploratory inquiry. Thus, we conceptualized the Triple-C Model as a critical way of augmenting traditional questions identifying relationship structures (e.g., single-item, self-identification questions). From this point of view, a global question would likely drive the identification of basic monogamy and nonmonogamy structures (e.g., polyamory, swinging, open-relationships), relying primarily on self-identification. The Triple-C would then help to clarify which of those monogamous or nonmonogamous structures represented healthy and vitalizing vs. maladaptive approaches to relationship structures. Thus, a healthy and vital CNM relationship would be characterized by not only identifying as an open relationship, but by also reporting low desire for monogamy for all relationship partners as well as high levels of communication and comfort with that structure. Similarly, a healthy and vital monogamous relationship would be characterized by not only identifying as monogamous, but also reporting high desire for monogamy by both partners and high levels of communication and comfort with that structure. The Triple-C model would go so far as to suggest that although non-traditional relationship structures with lower levels of mutual comfort, consent, and communication could be called nonmonogamous, they should not be identified as CNM, as that term was designed to imply high levels of mutual consent (and very likely high mutual comfort and communication as well).

Mutual Consent

First, in the context of the Triple-C model (the conceptual framework guiding this exploratory work), we would assert that the foundation of any relationship structure involves mutual consent, with both partners explicitly agreeing on that arrangement. Each relationship structure comes with its own set of agreed upon “rules,” whether it is sexual exclusivity, emotional exclusivity, or agreed upon contexts in which EDSA is allowable (i.e., at specific social events for swingers) or not allowable (i.e., not with mutual friends, not with ex-partners). Breaking relationship structure rules has been linked to poorer functioning both in monogamous couples (Mark, Janssen, & Milhausen, 2011) and in nonmonogamous couples (Hosking, 2013). Underscoring the importance of mutuality in these agreements, individuals from nonmonogamous relationships were more likely to perceive EDSA as cheating if they had not shared in the decision to open their relationships (Cohen, 2016). In addition, a lack of agreement about the degree of monogamy within a relationship has been linked to higher sexual jealousy (Parsons, Starks, Gamarel, & Grov, 2012) and to lower trust and commitment (Hoff, Beougher, Chakravarty, Darbes, & Neilands, 2010).

Ongoing Communication

Second, the Triple-C model asserts that communication around the relationship structure is another critical element of a healthy relationship structure. Communication has been robustly linked to both current and longitudinal changes in relationship satisfaction and stability (see Karney & Bradbury, 1995 for a review). Although essential to any relationship, we posit that communication specifically around EDSA might play a key role in nonmonogamous relationships, allowing couples to negotiate ongoing EDSA with various others while maintaining high levels of respect and consideration toward the feelings of each other. As mentioned above, even individuals in nonmonogamous relationships can perceive sexual betrayals (Hosking, 2013) and poor communication has been linked to between-partner discrepancies in perceptions of relationship structures (Hoff et al., 2010). Along these lines, secrecy surrounding EDSA has been linked to lower relationship satisfaction (Wagner, Remien, & Dieguez, 2000).

Mutual Comfort

Finally, the Triple-C model asserts that comfort with the agreed upon relationship structure is critical. If one partner felt coerced into agreeing to a nonmonogamous structure (potentially desiring monogamy but wanting to accommodate their partner’s desires for nonmonogamy) or simply felt less comfortable with a non-traditional relationship structure even after agreeing to it, then ongoing EDSA could very likely lead to hurt feelings and jealousy. Consistent with this, relationships marked by between-partner discrepancies in their perceptions of the degree of monogamy of their relationship experienced higher levels of sexual jealousy than truly consensual nonmon-ogamous relationships (Parsons et al., 2012). Even in relationships with agreed upon monogamous structures, one or both partners feeling less comfortable with that traditional structure could likely lead to conflict and discord over time (e.g., Denfeld, 1974).

Measuring/Operationalizing Nonmonogamy

A majority of previous studies have used single items, often giving descriptions of two basic relationship types (monogamy vs. some form of nonmonogamy) and asking respondents to choose the one that most directly applies to their relationships (e.g., Blasband & Peplau, 1985; Kurdek & Schmitt, 1986; Ramirez & Brown, 2010; Rodrigues, Lopes, & Pereira, 2016). Other studies diversify that operationalization by offering several separate response options (e.g., offering: swinging, polyamory and/or “open” relationships as separate answer choices; Conley et al., 2017; Moors, Conley, Edelstein, & Chopik, 2015). A few studies have expanded to using two separate items – typically one assessing engagement in EDSA (or multiple partners) and a second assessing relationship structures, thereby classifying individuals into 2–3 basic groups (Mogilski, Memering, Welling, & Shackelford, 2017; Parsons, Starks, Dubois, Grov, & Golub, 2013; Rodrigues, Lopes, & Smith, 2017; Swan & Thompson, 2016). Unfortunately, the specific items and answer choices offered notably varied across studies and across labs, making it difficult to integrate findings in this area. More to the point, such one or two-item classifications might inadvertently deemphasize the components of the Triple-C Model by failing to mention the levels of these three factors across the answer choices given. For example, by using only single items to assess relationship structures, open relationships with extremely high levels of mutual comfort, communication, and consent could inadvertently be grouped with open relationships in which one or more partners experience notable discomfort and disagreement with that structure. This would potentially confound the groups examined and would likely serve to obscure meaningful group differences that might otherwise have been obtained. Therefore, using only one or two item strategies to assess relationship structures runs the risk of oversimplifying these complex interpersonal arrangements. Of course, without concrete, data-driven guidelines for what types of relationship structures could be expected to exist within a national sample, it would be extremely difficult to craft a one or two item measure that would accurately represent the full range of possible relationship structures.

Correlates of Nonmonogamy

Links to Relationship Quality

Although the current work is fundamentally exploratory in nature, the Triple-C model guiding this study would suggest that EDSA in the context of a CNM structure could potentially invigorate romantic relationships, whereas EDSA within relationships with structures with lower levels of mutual comfort, consent, and communication or EDSA outside of the agreed-upon structure could represent a form of betrayal (Cohen, 2016), damaging those relationships. As previous work has often confounded these two types of nonmonogamy, the findings contrasting nonmonogamous to monogamous relationships has yielded mixed results. Thus, a majority of previous studies have failed to find differences in relationship quality (typically operationalized as relationship or sexual satisfaction in the primary relationship; Blasband & Peplau, 1985; Conley et al., 2017; Hosking, 2013; LaSala, 2004; Mogilski et al., 2017; Ramirez & Brown, 2010; Rodrigues et al., 2016, 2016; Rubin & Adams, 1986; Wagner et al., 2000; Wood et al., 2018). However, a few studies have found higher relationship quality in monogamous couples (Hoff et al., 2010; Kurdek & Schmitt, 1986; Swan & Thompson, 2016) and others found higher quality in the primary relationships of nonmonogamous couples (Bergstrand & Williams, 2000; Conley, Piemonte, Gusakova, & Rubin, 2018). Studies in which nonmonogamy groups were potentially confounded with monogamous relationships experiencing infidelity (e.g., the “partial” and “secrecy” groups of Wagner et al., 2000) or with individuals in not-yet exclusive dating relationships (e.g., Mogilski et al., 2017) yielded non-significant differences from monogamous relationships. However, recent studies using more precise assessment questions also yielded mixed results, with some showing non-significant differences from monogamous relationships (e.g., Conley et al., 2017; Wood et al., 2018) and another showing higher relationship quality in swingers than in monogamous couples but no significant differences between open relationships and monogamous relationships (Conley et al., 2018). Taken as a set, these seemingly inconsistent findings could be a direct result of not augmenting the assessment of self-identified relationship structures with the assessment of components of the Triple-C model. This could have led some studies to include greater numbers of relationships with less adaptive nonmonogamous structures (i.e., low comfort, communication, and consent) in their CNM groups. Our study further extended previous work by examining those relationships structures for differences not only on overall relationship quality (i.e., relationship satisfaction) but also on a broader set of relevant relationship characteristics including: relationship length and cohabitation status (to assess relationship demographics; e.g., Conley et al., 2018), dedication (to assess levels of commitment; e.g. Balzarini, Shumlich, Kuhot, & Campbell, 2018b), rates of affection & sexual activity (to assess intimate behavior), and sexual satisfaction (e.g., Muise, Laughton, Moors, & Impett, 2018). To build on this work, the current study included this broader range of scales, enabling a deeper characterization of the relationships represented within the various relationship structure groups.

Links to Background Factors and Individual Functioning

Previous work has linked nonmonogamy to non-heterosexual sexual orientations/identities (Balzarini et al., 2018a; Rubin, Moors, Matsick, & Conley, 2014; Solomon, Rothblum, & Balsam, 2005), unrestricted sociosexual orientations (i.e., more embracing of casual sex; Balzarini et al., 2018b; Mogilski et al., 2017; Rodrigues et al., 2016), and higher rates of drug use (Parsons et al., 2013), but has yielded mixed findings for attachment avoidance (Moors et al., 2015; Ramirez & Brown, 2010) and rates of sexually transmitted infections (STIs; e.g., Conley, Moors, Ziegler, & Karathanasis, 2012; Lehmiller, 2015; Platteau et al., 2017), and has failed to find differences on psychological distress (Wagner et al., 2000). Thus, although somewhat mixed, previous findings have suggested that nonmonogamous relationships might be associated with a certain degree of sexual and recreational disinhibition and a rejection of mainstream or hetero-normative values. The current study therefore sought to extend this work by exploring differences across the latent relationship structure groups extracted on a broader set of background factors and indices of individual functioning, including: sexual orientation identification, sociosexual orientation, sexual sensation seeking, attachment avoidance, psychological distress, loneliness (as it has been linked to infidelity in monogamous relationships; Pereira, Taysi, Orcan, & Fincham, 2014), heavy drinking, STIs, and condomless sex with new partners (capturing the dysregulated sexual activities that might lead to higher STI rates). To incorporate a rapidly expanding cultural phenomena that has exploded in popularity over the last 10 years, the current study also examined geosocial networking app use (GSN - smartphone apps that allow users to identify potential sexual and relationship partners nearby; i.e., Tinder, Grindr) as a salient marker of recent partner-seeking.

The Current Study

To build upon previous work, we sought to develop the Triple-C model as a conceptual framework for understanding both monogamous and nonmonogamous relationships, augmenting traditional single-item, self-identified assessments with factors critical to vital and healthy relationship structures. Toward that overarching goal, the study used an exploratory and data-driven approach to uncover the fundamental types of relationship structures that existed within a large and diverse sample of 1,658 online respondents in relationships.

Aim 1: Identifying Relationship Structure Groups with Latent Profile Analyses

In an effort to allow the diversity within our sample to determine the fundamental classes of relationship structure without introducing any biases or expectations, we used the exploratory technique of Latent Profile Analysis (LPA; McCutcheon, 1987). LPA identifies a parsimonious set of participant groups, helping to uncover fundamental underlying (latent) groups of people with respect to the indices examined, thereby clarifying how those factors might naturally co-occur within a population. Thus, our first aim was to conduct an LPA on items assessing the key elements of relationship structures.

Aim 2: Examining Possible Gender Differences in the LPAs

Although less widely examined, a handful of studies have highlighted that men are more willing to engage in CNM relationships (e.g., Denfeld, 1974; Moors et al., 2015) and are generally more embracing of casual sex (Simpson & Gangestad, 1991), whereas women are more likely to engage in monogamous relationships (Balzarini et al., 2018a). These results begin to suggest that men might have distinct orientations from women on relationship structures. We therefore ran secondary LPA analyses that extracted fundamental classes of relationship structures in men and women separately. These analyses allowed us to determine if the same classes of relationship structures emerged across men and women (suggesting gender invariance) or if unique classes emerged across genders (suggesting that men and women might view relationship structures from distinct perspectives).

Aim 3: Characterizing the Commitment Groups

After extracting the latent groups, we examined differences across those groups on a range of factors, characterizing the upstream and downstream correlates of these latent commitment groups.

Method

Participants

The online survey was open to all individuals in relationships (of any duration) who were 18 years of age or older. Respondents were recruited with the study title (“The Finding Partners in the 21^st Century Survey”) and the logistics of participation (voluntary, online, 25–30 min), yielding a final sample of 1,658 respondents. Respondents forming the current sample were recruited from Mechanical Turk (38.54%), ResearchMatch (34.02%), a psychology undergraduate pool (17.13%), psychology research websites (5.31%; e.g., socialpsych.org, University of Hanover), social websites (2.11%; e.g., Facebook), and via other methods (2.90%). As seen in Table 1, a majority of respondents were in their 20’s and 30’s, identified as White (78.65%) and female (69.78%), and were in some form of long-term relationships (together an average of 4.37 years, SD = 6.06 years, with 69.70% together ≥ 1 year and 82.13% together ≥ 6 months).

Table 1. Demographics of sample

Demographic Dimension Subsamples	N	% in full sample	% in males	% in females
Full sample	1658
Gender Identity
Male	485	29.25%
Female	1157	69.78%
Non-binary/Gender-fluid/Other	16	0.97%
Age
18-20 yrs	232	13.99%	10.72%	15.35%
21-24 yrs	401	24.19%	19.18%	26.26%
25-29 yrs	336	20.27%	20.00%	20.38%
30-39 yrs	382	23.04%	28.66%	20.72%
40+ yrs	301	18.15%	21.24%	16.88%
Race
White	1304	78.65%	77.94%	78.94%
Asian	143	8.62%	9.07%	8.44%
Black	95	5.73%	6.60%	5.37%
Other	105	6.33%	5.98%	6.48%
Ethnicity
Non-Latino/Hispanic	1485	89.57%	89.28%	89.68%
Latino/Hispanic	138	8.32%	8.87%	8.10%
Education
GED/HS diploma or less	146	8.81%	9.69%	8.44%
Some college/AA degree/Trade certificate	590	35.59%	35.88%	35.46%
Bachelor’s degree	587	35.40%	34.02%	35.98%
Graduate degree	332	20.02%	20.41%	19.86%
Income
0 to $29,999	411	24.79%	25.15%	24.64%
$30,000 to $59,999	505	30.46%	34.23%	28.90%
$60,000 to $89,999	310	18.70%	17.53%	19.18%
$90,000 or more	415	25.03%	22.89%	25.92%
Sexual Orientation
Heterosexual	987	59.33%	63.30%	57.97%
Heteroflexible	285	17.19%	10.72%	19.86%
Bisexual	190	11.46%	8.25%	12.79%
Homosexual	196	11.82%	17.73%	9.38%
Relationship Status
Casually dating (dating more than 1 person)	254	15.32%	18.97%	13.81%
In a committed relationship	509	30.70%	30.10%	30.95%
Long-term committed relationship	436	26.30%	21.86%	28.13%
Engaged	85	5.13%	4.33%	5.46%
Married	374	22.56%	24.74%	21.65%

Note: Sexual orientation was assessed with the 2-item Multidimensional Sexual Orientation Classification System (MSOCS: Legate & Rogge, 2019). Respondents were grouped into male and female groups based on their gender identities rather than their sex at birth (e.g., trans-males were grouped with cis-gendered males).

Measures

In creating the survey, we selected widely used and well-validated scales from the existing literature whenever possible. The items for the scales used in the study were presented predominantly on 6 or 7-point response scales (e.g., “Not at all” to “Completely”). Unless otherwise indicated, we used the response sets originally created for each scale and responses were averaged across the items of scales such that higher scores reflected higher levels of the constructs being assessed. All Cronbach alphas presented were calculated with the current sample.

LPA Predictors of Relationship Structure

The following scales were used to assess relationship structures, integrating both the more widespread use of a single item with a range of additional items representing components of the Triple-C model. All of these variables were entered into the LPAs to help identify latent classes of relationship structure.

Single-item Self-identified Relationship Structure

Building on the strategies commonly used in previous work, respondents’ self-identified relationship commitment structure was obtained using a single item, “What is the commitment structure of your relationship?” with the following response options: “Exclusive/monogamous (not engaging in sexual activity with others),” “Committed – Partly-open (only one of us engages in sexual activity with other people),” “Committed – Open relationship – Dyadic (we introduce 3rd partners into sexual activity but only together),” “Committed – Open relationship – Non-Dyadic (we both have sexual activity with other people),” and “Not committed.” Thus, this item assessed both how traditional (vs. non-traditional) and mutually consensual their relationship structures were.

Extradyadic Sexual Activity

The EDSA items were written to be: (1) clear and easily understandable, (2) neutrally worded (to avoid implying any sort of betrayal that might not apply to CNM relationships), and (3) were precise and behaviorally specific. We chose a 2-month timeframe as we felt it would be: (1) long enough to capture reasonable amounts of EDSA, (2) short enough to maximize accuracy in reporting, and (3) recent enough to have a strong impact on current functioning. Thus, respondents completed two items assessing their EDSA (“With how many individuals (other than your partner) have you had sexual activity in the last 2 months?” and “To your knowledge, with how many individuals (other than you) has your partner had sexual activity in the last 2 months?”).

Triple-C Model Components

Seven items were administered on a 6-point scale (“Not at all” to “Extremely”). These items were generated by the research team to be clear, straightforward, and easily understandable to maximize their relevance for subjects. We therefore developed pairs of items to assess the components of the Triple-C model for respondents and their partners.

Desire for Monogamy

Two items assessed desire for monogamy (“How much do you (does your partner) believe in monogamy?” and “How much do you (does your partner) desire a monogamous relationship?”; low levels for both partners further suggesting high mutual consent).

Distress at EDSA

Two items assessed distress at EDSA (“How upset would you be if you knew that your partner was engaging in sexual activity with other people?” and “How upset would your partner be if he/she knew the full extent of your sexual activity with others?”; low levels for both partners suggesting high mutual comfort).

Awareness of EDSA

Two items assessed awareness of EDSA (“How aware are you of your partner’s sexual activity with others?” and “How aware is your partner of your sexual activity with others?”; high levels suggest high levels of comfort and communication).

Communication about EDSA

One item assessed communication about EDSA (“How much do you and your partner actively communicate about your sexual activity with others?”).

Proximal Relationship Factors

To provide insights into the general nature of the various relationships within these different classes of relationship structures, respondents provided some basic relationship demographics (length, stage, cohabitation) and completed the following scales.

Relationship Satisfaction

Respondents completed the 4-item Couples Satisfaction Index (CSI-4; Funk & Rogge, 2007; α = .94) assessing global evaluations of relationship quality (e.g., “In general, how satisfied are you with your relationship?”). Responses were summed using the validated scoring system so that higher scores reflected higher relationship satisfaction.

Dedication

Respondents completed 4 dedication items from the Commitment Inventory (CI; Stanley & Markman, 1992; α = .88), assessing a long-term commitment to a current partner (e.g., “I want this relationship to stay strong no matter what rough times we encounter”).

Frequency of Physical Affection

Respondents completed 4 items on the frequency of physical affection in the past 2 weeks (cuddle, hold one another, hug, deep kiss/make out; α  = .93).

Frequency of Sexual Activity

Respondents completed 5 items assessing frequency of sexual activity in the past 2 weeks (have oral sex, have anal sex, have vaginal sex, climax with each other, masturbate together; α = .82).

Sexual Satisfaction

Respondents completed the 6-item sexual satisfaction subscale of the Quality of Sex Inventory (QSI; Shaw & Rogge, 2016; α  = .97), assessing global evaluations of respondents’ sexual relationships with their partners (e.g., “I am happy with the quality of sexual activity in our relationship”).

Background Factors

To integrate previous work and provide a more comprehensive understanding of the various types of individuals that gravitate to different relationship structures, respondents provided basic demographics (age, race/ethnicity, income, education, gender) and completed the following scales.

Sexual Orientation

Respondents reported their sexual orientation by completing the 2-item Multivariate Sexual Orientation Classification System (MSOCS; Legate & Rogge, 2019), a system that integrates sexual identity, attraction, and behavior to classify people into 4 basic sexual orientation classes (validated by LPA, see Legate & Rogge, 2019). The items assess degree of divergence from heterosexuality (“Please indicate the sexual orientation with which you identify” on a “0 – exclusively heterosexual behavior” to “6 – exclusively homosexual behavior” scale) and level of sexual attraction to same-sex partners (“How much are you sexually attracted to [men/women – matching subject’s own gender identity]?”).

Sociosexual Orientation

Respondents completed the 9-item Revised Sociosexual Orientation Inventory (SOI-R; Penke & Asendorpf, 2008; α  = .87) to indicate their level of desire for, comfort with, and experience with casual sex (e.g., “I can imagine myself being comfortable and enjoying casual sex with different partners”).

Sexual Sensation Seeking

Respondents completed 9 internally-consistent items from the Revised Sexual Sensation Seeking Scale (R-SSSS; Kalichman & Rompa, 1995; α = .87), assessing interest in sexual exploration and risk-taking (e.g., “I like wild uninhibited sexual encounters”).

Attachment Avoidance

Respondents completed the 6 most informative attachment avoidance items of the Experiences in Close Relationships – Revised scale (ECR-R; Fraley, Waller, & Brennan, 2000; α = .83; e.g., “I prefer not to be too close to romantic partners”).

Outcomes – Psychological

To examine links between different fundamental relationship structures and current emotional functioning, respondents completed the following scales.

Need Satisfaction

Respondents completed the 21-item Basic Psychological Need Satisfaction Scale (BPNSS; Gagné, Ryan, & Bargmann, 2003; α = .89), assessing the satisfaction of autonomy needs (e.g., “I feel like I am free to decide for myself how to live my life”), relatedness needs (e.g., “People in my life care about me”), and competency needs (e.g., “Often, I do not feel very competent” – reversed)

Loneliness

Respondents completed the 8-item version UCLA Loneliness Scale (ULS-8; Hays & DiMatteo, 1987; α  = .93; e.g., “How often do you feel isolated from others”).

Psychological Distress

Respondents completed eight items of the distress subscale of the Mood and Anxiety Symptom Questionnaire (MASQ; Watson et al., 1995; α  = .97; e.g., “In the last 2 weeks, I … felt depressed, felt sad, felt hopeless”).

Outcomes – Behavioral

Finally, to examine links between basic relationship structures and risk behaviors and online partner-seeking, respondents completed the following items.

Geosocial Networking App Use

Respondents indicated if they had recently used any GSN apps, “IN THE LAST 2 MONTHS, How often did you use geosocial apps (apps for smart devices that use your location to find other users nearby e.g., Tinder, Grindr)?”, creating a dichotomous variable flagging the presence or absence of recent GSN use to be used in our analyses.

Heavy Drinking/drug Use

Respondents completed 10 items assessing frequency of heavy drinking and Drug use (e.g., “In the last 2 months, how often did you: get drunk, use illegal drugs, have a hangover due to drinking and/or drug use, get sick due to drinking and/or drug use, miss a class/work due to drinking and/or drug use”; α = .89).

Condomless Sex

A subset (n = 443) of respondents reporting meeting new individuals via recent social activity (i.e., geosocial app use, dating website use, and/or offline social activities – e.g., going to bars, parties, clubs) completed items assessing recent condomless intercourse with those individuals (“With how many people from [specific venue] did you engage in the following behaviors: oral sex without a condom, vaginal intercourse without a condom, anal intercourse without a condom”).

Sexually Transmitted Infections

Respondents indicated all STIs they had ever contracted out of 10 common STIs (“Please indicate the sexually transmitted infections you have contracted: pubic lice, scabies, gonorrhea, chlamydia, HPV, HPV with genital warts, hepatitis, vaginitis (including trichomoniasis, gardnerella, & candidiasis), herpes, syphilis”). Given the low rates of endorsement, responses on these items were converted to a single dichotomous variable contrasting individuals reporting no experience with STIs from those reporting one or more.

Procedure

Data were collected from March – July 2016 as part of a larger, 25–30 min, online survey securely hosted on SurveyGizmo.com that was open to all adults (18+ years old). Respondents were given individualized feedback on a number of factors within the survey as the primary incentive. Respondents recruited from Mturk were given $0.40 of amazon.com store credit and students from the undergraduate subject pool were given extra credit as additional incentives.

Analytic Strategy

Latent Profile Analyses

LPA is a form of latent mixture modeling conducted within a Structural Equation Modeling framework. The primary goal of exploratory LPA is to identify a parsimonious set of participant classes on a set of variables (McCutcheon, 1987), clarifying the smallest number of groups that meaningfully distinguish respondents from one another. This is achieved by running a series of LPAs with increasing numbers of specified classes and then identifying the most parsimonious model (i.e., accounting for the greatest amount of between person differences with the fewest numbers of classes). To identify a parsimonious model, one must find a number of classes that optimizes: (1) model fit indices (i.e., lower AIC and BIC indices, significant Lo-Mendel-Ruben and Bootstrapping Likelihood Ratio Tests), (2) the entropy of each resulting model (the average probability of each participant belonging to the class to which he or she was ultimately assigned, with high values – preferably above .9 – representing clarity of classification), and (3) the interpretability and relevance of each solution (favoring solutions lacking exceedingly small classes that would be less likely to replicate; see Calzo, 2014 for a deeper discussion of balancing these criteria). Toward this end, a series of LPA models on the indicators of relationship structures were run in the current data using MPlus, version 7.11 (Muthen & Muthen, 2012) specifying from 2 to 7 classes. To ensure that the latent class solution was stable across gender, secondary LPAs were run separately in men and women (see Table 2).

Table 2. Model fit across latent profile analyses specifying from 1 to 5 latent classes

Classes Specified	# of free parameters	Log- likelihood	AIC	BIC	BIC N-adj.	LMR-LRT p	BLRT p	Entropy	# classes with n<40
FULL SAMPLE
2	34	-23715	47498	47682	47574	<.0001	<.0001	0.974	0
3	48	-22943	45982	46242	46089	0.003	<.0001	0.954	0
4	62	-22273	44669	45005	44808	<.0001	<.0001	0.978	0
5	76	-21830	43811	44223	43981	0.003	<.0001	0.956	0
6	90	-21295	42770	43258	42972	0.098	<.0001	0.972	0
7	104	-20975	42158	42722	42391	0.068	<.0001	0.97	0
IN 485 MEN
2	34	-7098	14264	14406	14298	0.0664	<.0001	0.964	0
3	48	-6843	13782	13983	13830	0.0007	<.0001	0.946	0
4	62	-6630	13384	13643	13446	0.0008	<.0001	0.972	0
5	76	-6492	13136	13454	13213	0.0139	<.0001	0.963	0
6	90	-6360	12901	13277	12992	0.0728	<.0001	0.97	0
7	104	-6233	12673	13108	12778	0.0109	<.0001	0.974	0
IN 1173 WOMEN^1
2	34	-16539	33145	33317	33209	<.0001	<.0001	0.978	0
3	48	-16015	32127	32370	32217	0.0001	<.0001	0.964	0
4	62	-15554	31232	31546	31349	0.1504	<.0001	0.965	0
5	76	-15217	30586	30972	30730	<.0001	<.0001	0.978	0
6	90	-14822	29824	30280	29994	<.0001	<.0001	0.976	0
7	104	-14608	29425	29952	29622	0.0018	<.0001	0.974	0

Note: The table presents the fit statistics from LPAs run in the full sample (top section), in men only (middle section), and in women¹ only (bottom section), demonstrating that the fit indices continued to suggest a 5 class solution in both major gender groups. The solution selected to have the most parsimonious fit has been bolded to ease interpretation. AIC = Aikaike’s information criterion, BIC = Bayesian information criterion, N-adj. = BIC adjusted for sample size (lower numbers on the AIC and BIC indices suggest better fit), LMR-LRT p = Lo-Mendel-Ruben likelihood ratio test significance level, BLRT p = Bootstrap Likelihood Ratio Test significance level (a significant LMR-LRT or BLRT suggests that the current solution accounts for significantly greater amounts of between-person differences than the solution with one fewer classes specified). Entropy = the average probability of each participant belonging to the class to which they were ultimately assigned, thereby representing clarity of classification for each solution (higher numbers suggest a better fit with numbers above .9 suggesting excellent fit). ¹ So as to include all respondents in one of the two groups, the 16 transgender/non-binary/other respondents were combined with the women when running the final set of LPAs above.

Developing a Parsimonious Algorithm

To develop a straightforward method of reproducing the latent groups identified in the LPAs and to gain some initial clarity on how those groups differ from one another, we ran Decision Tree analyses (using the DecisionTreeClassifier from the Python scikit-learn package; Pedregosa et al., 2012) to identify the smallest number of items necessary to identify these groups in future studies.

Characterizing Latent Groups

Although our primary goal was to use the Triple-C model (in a large sample with person-centered analyses) to identify fundamental relationship structures, another major goal of this study was to integrate much of the previous work on correlates of nonmonogamy and extend it by examining the fundamental relationship structures identified on a broader set of dimensions. Thus, we wanted to be able to provide readers with a sense of the demographics and backgrounds of the individuals in those various latent classes, the nature of the relationships in those latent classes, and the types of sexually-related behavior exhibited by individuals in those classes. To examine differences among the final LPA groups (Tables 3 and 4), we ran a series of separate univariate ANOVA and χ² analyses (allowing for maximum inclusion of subjects in these analyses) to test for omnibus differences across the groups on the variables that we used to create those groups as well as a broad set of constructs examining relationship and individual functioning across those relationship structure groups. Following analyses treating them as continuous variables, a handful of continuous variables were dichotomized using cutscores with real-world meanings to increase the interpretability of their findings. Given the planned number of contrasts, we used a threshold of p < .001 to identify significant differences and a threshold of p < .01 to identify marginal differences for all of the omnibus tests presented. This served to both constrain the experiment-wide alpha error levels and to focus the results on the more robust differences. All omnibus tests revealing significant differences were then followed up with post-hoc pairwise comparisons using Tukey HSD (for the ANOVAs) or Bonferroni corrections (for the χ² tests) to further constrain experiment-wide alpha levels. To place the observed group differences on a common metric and provide readers with a concrete sense of the relative magnitude of the effects presented, we also calculated Cohen’s d statistics contrasting the two groups demonstrating the largest difference on each variable.

Table 3. Characterizing the latent commitment classes

					Latent Classes
Class of Constructs					Monogamous		Nonmonogamous
Specific Constructs Examined				Full Sample	MONOG - min EDSA	MONOG - low EDSA	OPEN CNM	PART-OPEN	ONE SIDED	Omnibus test of differences across classes	Effect Size of Greatest Pairwise Diff (Cohen’s d’s)
Number of respondents:				1,658	629	496	128	218	187
LATENT PROFILE ANALYSIS PREDICTORS						Within-Group Percentages or Means
Global Structure Question
R1: Monogamous				72.26%	95.7%^A	87.00%^B	7.70%^E	49.10%^C	26.40%^D	χ ^2(4)=758*	3.1
R2: Primarily one partner engages in EDSA				10.13%	1.00%^D	5.90%^C	11.70%^BC	18.20%^B	40.90%^A	χ ^2(4)=299*	1.4
R3: Open relationship - EDSA together				4.40%	0.30%^C	4.70%^BC	21.70%^A	8.60%^B	0.70%^C	χ ^2(4)=143*	0.93
R4: Open relationship - EDSA separately				7.30%	0.20%^D	1.90%^C	53.20%^A	12.20%^B	8.50%^B	χ ^2(4)=461*	1.88
R5: Not committed				5.91%	2.80%^C	0.50%^D	5.70%^BC	11.90%^B	23.50%^A	χ ^2(4)=149*	0.89
Recent EDSA
Reported own EDSA in last 2 months				18.79%	3.30%^D	13.70%^C	63.30%^A	34.20%^B	60.70%^A	χ ^2(4)=400*	1.77
Reported partner EDSA in last 2 months				12.96%	2.20%^D	8.70%^C	58.50%^A	27.20%^B	28.00%^B	χ ^2(4)=300*	1.74
Triple-C Components
CN	CF	CM
x	x		Own desire for monogamy	4.72	5.28^A	5.23^A	2.39^D	4.24^B	3.66^C	F(4,1651)=327*	3.28
x	x		Partner’s desire for monogamy	4.96	5.52^A	5.54^A	2.36^D	4.36^B	3.99^C	F(4,1646)=421*	4.16
	x		Own distress at partner’s EDSA	3.69	3.45^B	4.76^A	1.65^C	3.53^B	3.18^B	F(4,1638)=78*	1.83
	x		Partner’s distress at own EDSA	2.34	1.80^C	2.65^B	1.31^D	2.79^B	3.45^A	F(4,1640)=49*	1.34
	x	x	Own awareness of partner’s EDSA	3.16	1.08^D	5.38^A	5.47^A	3.51^B	2.18^C	F(4,1642)=1553*	7.46
	x	x	Partner’s awareness of own EDSA	3.16	1.03^D	5.68^A	5.71^A	3.58^B	1.32^C	F(4,1648)=11699*	19.82
		x	Active communication about EDSA	2.72	1.33^E	4.07^B	5.01^A	3.05^C	1.79^D	F(4,1641)=363*	3.33
PROXIMAL RELATIONSHIP FACTORS
Less than 1 year				30.30%	22.81%^C	31.65%^AB	24.41%^BC	41.94%^A	42.39%^A	χ ^2(4)=46*	0.38
1 to 5 years				44.00%	46.60%^A	48.00%^A	37.80%^AB	41.00%^AB	32.10%^B	χ ^2(4)=18*	0.29
More than 5 years				25.70%	30.60%^A	20.40%^B	37.80%^A	17.10%^B	25.50%^AB	χ ^2(4)=34*	0.48
Living together				43.68%	50.56%^A	43.43%^AB	54.33%^A	26.27%^B	34.22%^B	χ ^2(4)=51*	0.59
Relationship satisfaction				15.63	16.03^A	16.92^A	16.67^A	13.96^B	12.09^C	F(4,1651)=61*	1.25
Dedicated (top 3 of 6 answers)				74.60%	83.50%^A	84.20%^A	75.80%^A	54.60%^B	42.20%^B	χ ^2(4)=199*	0.93
Low Affection (0-4x in last 2 weeks)				28.80%	29.90%^B	22.00%^C	18.00%^C	33.50%^AB	45.50%^A	χ ^2(4)=47*	0.59
Frequency of sexual activity w partner				1.18	1.01^C	1.30^AB	1.28^ABC	1.50^A	1.05^BC	F(4,1647)=10*	0.44
High sexual satisfaction (top 2 of 6)				45.10%	46.40%^A	54.50%^A	49.60%^A	29.40%^B	31.20%^B	χ ^2(4)=56*	0.48

Note: MONOG – min EDSA = Monogamous relationships with minimal extra dyadic sexual activity; MONOG – low EDSA = Monogamous with low amounts of recent EDSA; CNM OPEN = Open relationships that could be classified as consensual nonmonogamy; PART – OPEN = Partially open relationships; ONE SIDED = Relationships with Primarily One-Sided EDSA; R1-R5 = the response options offered for the question asking participants to identify the relationship commitment structure of their relationship. CN = Consent, CF = Comfort, CM = Communication. Given the number of planned contrasts, a p = .001 significance threshold was used for the omnibus tests and a p = .005 threshold for the subsequent pairwise comparisons to help constrain the experiment-wide alpha level. Different superscripted letters indicate significant differences between classes emerging from the pairwise comparisons. Cohen’s d’s were calculated for the largest difference between group means on each variable (row) to give an estimate of a standardized effect size. * = p < .001, † = p < .005. The highest means or percentages in each row have been bolded to facilitate interpretation of the results.

Table 4. Differences across latent commitment classes on background, demographic and outcome factors

		Latent Classes
Class of Constructs		Monogamous		Nonmonogamous
Specific Constructs Examined	Full Sample	MONOG - min EDSA	MONOG - low EDSA	CNM OPEN	PART-OPEN	ONE SIDED	Omnibus test of differences across classes	Effect Size of Greatest Pairwise Diff (Cohen’s d’s)
Demographic Factors	Within-Group Percentages or Means
Male	29.25%	23.05%^B	29.03%^AB	35.94%^A	37.16%^A	36.90%^A	χ^2(4)=26*	0.28
Age	30.00	31.01^AB	27.79^C	32.57^A	29.26^BC	31.56^AB	F(4,1647)=11*	0.52
Years of education	15.20	15.52^A	14.71^B	15.39^A	15.22^AB	15.30^AB	F(4,1650)=8.2*	0.33
White	78.65%	81.56%^AB	75.00%^B	88.28%^A	75.23%^B	75.94%^B	χ ^2(4)=17*	0.32
Background Factors
Sexual Orientation (MSOCS)
Heterosexual	60%	70.27%^A	64.72%^A	24.22%^C	43.12%^B	52.94%^B	χ ^2(4)=130*	0.58
Heteroflexible	17%	11.76%^C	14.52%^BC	32.03%^A	22.48%^AB	26.20%^A	χ ^2(4)=50.3*	0.35
Bisexual	11%	7.79% ^B	9.68% ^B	28.13%^A	16.97%^AB	10.70%^B	χ ^2(4)=51.6*	0.36
Homosexual	12%	10.17%	11.09%	15.63%	17.43%	10.16%	χ ^2(4)=10.7	0.16
High sociosexual orientation (top 4 answers out of 9)	15.30%	7.30%^C	9.30%^C	42.50%^A	20.90%^B	32.80%^A	χ ^2(4)=167*	0.84
High sexual sensation seeking (top 3 answers of 6)	24.30%	16.00%^B	21.30%^AB	43.70%^A	29.90%^A	40.70%^A	χ^2(4)=81*	0.53
Attachment Avoidance	2.50	2.37^B	2.30^B	2.25^B	3.01^A	3.06^A	F(4,1596)=38*	0.84
Outcomes – Psychological
Loneliness	2.48	2.31^B	2.39^B	2.34^B	3.04^A	2.79^A	F(4,1595)=21*	0.65
Psychological distress	2.37	2.29^B	2.35^B	2.24^B	2.70^A	2.44^AB	F(4,1563)=4.3†	0.34
Outcomes – Behavioral
Geosocial networking app use	22.44%	9.22%^C	18.15%^B	41.41%^A	43.12%^A	41.18%^A	χ ^2(4)=186*	0.84
Heavy drinking/drug use (1-2x or higher on average across all heavy use/impairment items in last 2 mo)	21.20%	12.70%^C	21.60%^B	15.70%^BC	37.50%^A	34.10%^A	χ ^2(4)=79*	0.57
Condomless sex with new partners in last 2 mo (among a subset of 443 of the Ps actively seeking new partners)	50.79%	30.99%^B	38.78%^B	52.56%^AB	65.74%^A	60.23%^A	χ ^2(4)=30*	0.66
Sexually Transmitted Infections - Lifetime	25.38%	21.29%^B	20.59%^B	37.19%^A	34.76%^A	33.71%^A	χ ^2(4)=37*	0.72

Note: MONOG – min EDSA = Monogamous relationships with minimal extra dyadic sexual activity; MONOG – low EDSA = Monogamous with low amounts of recent EDSA; OPEN – CNM = Open relationships that could be classified as consensual nonmonogamy; PART – OPEN = Partially open relationships; ONE SIDED = Relationships with Primarily One Sided EDSA; Ps = Participants. MSOCS – the Multivariate Sexual Orientation Classification System (Legate & Rogge, 2019). Given the number of planned contrasts, a p = .001 significance threshold was used for the omnibus tests and a p = .005 threshold for the subsequent pairwise comparisons to help constrain the experiment-wide alpha level. Different superscripted letters indicate significant differences between classes emerging from the pairwise comparisons. Cohen’s d’s were calculated for the largest difference between group means on each variable (row) to give an estimate of a standardized effect size. * = p < .001, † = p < .005. The highest means or percentages in each row have been bolded to facilitate interpretation of the results.

Results

Aims 1 & 2: Identifying Latent Classes of Relationship Commitment

Aim 1: Latent Profile Analyses

In the analyses run in the full sample (first section of Table 2), a key indicator of model fit (the adjusted Lo-Mendel-Ruben Likelihood Ratio Test; LMR-LRT) suggested that a 5-class solution accounted for significantly greater between-subject differences than a 4-class solution, whereas solutions with larger numbers of classes failed to account for greater variance. As all of the solutions run in the full sample demonstrated appropriately high levels of entropy, these results would suggest that a 5-class solution was the most parsimonious.

Aim 2: Examining Possible Gender Differences

As shown in the second section of Table 2, secondary LPAs run in male respondents alone yielded a comparable pattern of fit, once again suggesting that a 5-class solution represented the most parsimonious fit to the data. The fit indices from secondary LPAs run in women deviated from this pattern, as the LMR-LRT suggested that a solution with as few as three classes might offer reasonable fit (third section of Table 2). However, a 5-class solution still demonstrated excellent fit in women and was supported by the remaining indices. More importantly, the five classes extracted in women were virtually identical to the 5 classes extracted in men and the 5 classes extracted in the full sample. Taken together, these results would suggest that the final 5-class solution was stable across men and women.

Describing Latent Classes of Relationship Commitment

As shown in the top half of Table 3 and in Figure 1a and b, the latent groups demonstrated notable differences on the predictor variables from which they were created.

Figure 1. Latent relationship structures extracted by LPA. S = subjects referring to themselves; P = subjects’ perceptions of their romantic partners; Monog = Monogamy; EDSA = extra dyadic sexual activity; MONOG – min EDSA = Monogamous relationships with minimal amounts of recent extra-dyadic sexual activity; MONOG – low EDSA = Monogamous relationships with low amounts of recent EDSA; OPEN – CNM = Open relationships that could be classified as consensual nonmonogamy.

Monogamous with Minimal Recent EDSA

The most common class (629 individuals; 37.93%) represented traditionally monogamous relationships: a class made up almost entirely of individuals identifying as being in monogamous relationships and reporting the lowest rates of recent EDSA, with correspondingly low rates of EDSA communication and awareness and reporting the highest levels of desire for monogamy.

Monogamous with Low Rates of Recent EDSA

The second most common class (496 individuals; 29.91%) represented monogamous relationships that might have recently experienced and/or discussed EDSA: the vast majority identifying as monogamous, with slightly higher proportions reporting recent EDSA as well as some of the highest levels of own and partner’s awareness of EDSA and the highest level of own distress at partner’s EDSA. Taking the first two groups together, a majority of the sample (67.85%) was classified in one of these two more traditional monogamous relationship structures.

Open-CNM Relationships

A total of 128 individuals (7.72%) were classified into a group representing Consensual nonmonogamy: a class with the highest rates of identifying as being in some form of a consensual open relationship (74.90% across the two possible responses, suggesting mutual consent), the highest proportions of recent EDSA (58.50–63.30%), with the lowest levels of desire for monogamy, lowest levels of distress at EDSA (i.e., high comfort), along with the highest levels of EDSA awareness and communication. Thus, even without placing any constraints or preconceptions on the analyses, a class of relationships emerged that reflect the true spirit of CNM within the Triple-C model in which EDSA is mutual and is handled openly and honestly.

Partially Open Relationships

Another group of 218 (13.14%) individuals were classified into a group representing more mixed attitudes and less clear arrangements regarding monogamy and EDSA: primarily identifying as being in monogamous relationships (49.10%), with fairly high proportions reporting recent EDSA (27.20–34.20%) in the context of reasonably high levels of desire for monogamy. This group showed only moderate amounts of partner awareness and communication surrounding EDSA as well as intermediate levels of distress at EDSA, suggesting notably mixed and potentially conflicting feelings and attitudes surrounding the EDSA in their relationships.

Relationships with Primarily One-sided EDSA

A total of 187 individuals (11.27%) were classified into a group representing individuals willing to engage in EDSA, despite knowing that their romantic partners would be upset over that behavior. This group reported the highest levels of being in a relationship in which primarily one partner engages in EDSA (40.90%), some of the highest rates of own recent EDSA (60.70%), and the highest level of partner’s distress at one’s own EDSA. Notably, these individuals seem to limit their partners’ knowledge of their EDSA, reporting the lowest levels of communication. Thus, the LPA results revealed that over two thirds of the nonmonogamous relationships in the sample fell into one of these final two groups in which desires for EDSA are in conflict with desires for monogamy in one or both partners.

The COMMIT4: A 4-item Algorithm

As shown in Figure 2, the decision tree analyses identified a 4-item algorithm that could reproduce the latent classes extracted with 93% accuracy. These questions included: the global question about relationship structure, the question about a partner’s knowledge of a respondent’s EDSA, and the two questions about a partner’s desire for monogamy. This algorithm helps to clarify that the Monogamous-with-Minimal-EDSA relationships primarily answered “not applicable” to the partner’s knowledge of your EDSA question and “monogamous” to the global relationship structure question. In contrast, although 87.00% of the Monogamous-with-Low-EDSA relationships also identified as monogamous on the global question, they were particularly defined by the respondents’ partners having very high awareness of the respondents’ EDSA and strong desire for monogamy. Turning to the nonmonogamous groups, the Primarily-One-Sided-EDSA relationships represent nonmonogamous relationships in which partners have little to no awareness of the respondents’ own EDSA, whereas the Partially-Open relationships represent relationships in which romantic partners only have moderate knowledge of the respondents’ EDSA. Finally, although 74.90% of the Open-CNM relationships identified with one of the two open-relationship options on the global relationship structure question, they were particularly defined as relationships in which partners are very aware of respondents’ EDSA and have little to no interest in monogamy.

Figure 2. The COMMIT4: a four-item reconstruction of latent classes derived from decision tree analyses. Decision tree analyses (a form of machine learning) were conducted using the DecisionTreeClassifier from the scikit-learn package (Pedregosa et al., 2012). The tree was constrained to five leaf nodes corresponding to the five desired latent classes (See the methods section for the full range of responses offered for question 2). The COMMIT4 classification system was derived via a machine learning decision tree algorithm to represent the most parsimonious method of recreating the 5 latent classes with the fewest items. It was therefore not intended to represent substantive findings (nor should it be substantively interpreted), but instead represents a practical application for future research.

Aim 3: Characterizing Latent Classes of Relationship Commitment

Monogamous Relationships

To further explore the distinguishing characteristics of the latent groups, we systematically characterized those groups across a diverse array of demographic, relationship, and individual factors. The Monogamous-with-Minimal-EDSA and the Monogamous-with-Low-EDSA groups were similar in that they both tended to have relatively healthy relationships: reporting some of the highest levels of relationship satisfaction, some of the highest proportions of dedicated respondents, and some of the highest proportions with high sexual satisfaction (bottom half of Table 3). Respondents in these groups also reported: some of the lowest levels of inadequate need satisfaction, loneliness, and psychological distress, some of the most restricted sociosexuality, and the lowest levels of sexual sensation seeking, suggesting fairly restrained and mainstream attitudes toward casual sex (Table 4). Taken together, these results suggest that individuals in the two groups of monogamous relationship structures were comfortable with the monogamous relationship structure of their relationships, reporting fairly high individual and relationship functioning within those relationships.

Respondents in these two groups primarily differed from one another, in that the individuals in the Monogamous-with-Low-EDSA group were more likely to: (1) report recent EDSA, (2) report greater awareness of EDSA, (3) have higher likelihood of relationships under a year in length and lower likelihood of relationships over 5 years, (4) involve younger individuals, and (5) report heavying drinking and GSN use. These results suggest that the second class of monogamous relationships might represent individuals who happen to be in monogamous relationships in their earlier stages. This could help to explain the slightly higher levels of relationship quality reported as well as the higher levels of communication around EDSA, as these individuals might have more recently discussed and decided upon their relationship structures.

Nonmonogamous Relationships

Individuals in the three classes of nonmonogamous relationships were similar to one another in that: (1) the partners reporting on the relationships were more likely to be male, (2) they were more likely to report unrestricted sociosexual orientations and (3) high levels of sexual sensation seeking, (4) they were more likely to report GSN app use within the last 2 months (i.e., actively seeking new partners), and (5) report ever having contracted an STI. Despite sharing similar permissive attitudes toward sex and similar levels of recent partner-seeking and sexual risk-taking, relationships in these three groups demonstrated marked differences on individual and relationship functioning.

Open-CNM Relationships

Respondents in this group tended to be in fairly long-term relationships and had the highest proportion of individuals living with their partners. This group also had the highest proportions of heteroflexible (identifying as primarily heterosexual but open to sexual activity with same sex partners; see Legate & Rogge, 2019) and bisexual respondents, suggesting that individuals in the LGBT community might be more comfortable embracing non-traditional structures. Individuals in these relationships also reported comparably high levels of relationship quality and individual functioning with respect to relationship satisfaction, dedication, affection, sexual satisfaction, inadequate need satisfaction, loneliness, and psychological distress. Taken together, these findings begin to suggest that the Open-CNM relationships in the current sample were notably healthy, robust and long-lived.

Other Nonmonogamous Structures

The two remaining nonmonogamous classes of relationships (partially open and one-sided) were similar to one another in that respondents in both classes reported the highest proportions of younger relationships, lower levels of individuals dedicated to their relationship, low levels of affection, low proportions of high sexual satisfaction, and the highest rates of condomless sex with new partners in the last two months. The class of Primarily-One-Sided-EDSA relationships also had the highest proportions of individuals significantly dissatisfied with their relationships (60%; nearly three times higher than the proportions in the monogamous or Open-CNM groups) and the highest rates of individuals with low sexual satisfaction. As seen in Table 4, the classes of Partially-Open relationships and Primarily-One-Sided-EDSA relationships were also similar in that they reported some of the highest levels of attachment avoidance and notably poorer psychological functioning, with higher proportions of individuals with inadequate need satisfaction and the highest levels of psychological distress and loneliness. Taken together, these results suggest that when somewhat avoidant individuals find themselves in notably unhappy and discordant relationships, they might turn to EDSA (among other dysregulated behaviors like heavy drinking) as a means of coping. Further adding to their relationship distress, these individuals engage in EDSA even if that necessitates doing so without their partners’ full knowledge or consent, and doing so despite knowing how much it would upset that partner. Alternatively, these results could suggest that one-sided EDSA might be particularly toxic to romantic relationships, rapidly eroding relationship quality.

Discussion

To our knowledge, this study represents the first published work to use an exploratory data-driven approach to examine latent classes of relationship structure. We identified two forms of more traditional monogamous relationships and three discreet forms of nonmonogamous relationships. Consistent with the Triple-C Model, the results highlighted mutual consent, comfort, and communication surrounding commitment and EDSA as key factors that could promote relationship health or relationship discord. As a majority of previous studies on CNM failed to comprehensively assess the components of the Triple-C model, the groups compared in those studies likely represent mixtures of the classes uncovered in the current study (Table 3; Figure 1a & b). This helps explain the largely non-significant (Conley et al., 2017; Rodrigues et al., 2016, 2016) and mixed findings (Bergstrand & Williams, 2000; Hoff et al., 2010; Kurdek & Schmitt, 1986; Swan & Thompson, 2016) in the literature. The current study therefore sheds key conceptual insights into when nonmonogamy can be damaging or beneficial to relationships. The results highlight that EDSA in the absence of mutual consent, comfort, and communication could easily come to represent a form of betrayal or cheating.

Implications

Monogamy Still Exists

A majority of respondents (1,125 or 67.85%) were classified into one of the monogamous groups, highlighting that monogamy still remains the prevalent relationship structure in the United States. In fact, individuals in the largest monogamous group (n = 629) also reported fairly low rates of EDSA in the last 2 months (3.30% own EDSA; 2.20% partners’ EDSA). Thus, over 96% of individuals in that largest group identified as monogamous and reported no recent EDSA – remaining true to that monogamous structure. This level is comparable to 12-month prevalence estimates of infidelity within married individuals from national samples (e.g., Whisman, Gordon, & Chatav, 2007), and suggests that a majority of individuals tend to honor their monogamous commitment agreements. Of course, the current results are based on a sample of convenience in which individuals with alternative lifestyles were purposefully oversampled (to facilitate the analyses). Thus, future work will need to examine these latent classes in nationally representative samples to clarify their prevalence and incidence rates.

Not All Monogamous Relationships are the Same

Unexpectedly, two distinct classes of monogamous couples emerged in the latent profile analyses. These two groups primarily differed in the amount of recent EDSA and in the level of awareness of EDSA. Thus, the LPA results served to identify a class of monogamous relationships at more than four times the risk of having recently engaged in EDSA (comparing MONOG-low EDSA to MONOG-min EDSA) despite that relationship structure. Although future research is needed to clarify the nature of that second monogamy class, the results suggested that those tended to be younger relationships in earlier stages, raising the possibility that those couples had more recently transitioned to monogamy. However, the results also revealed that the individuals in that class were less likely to be living with their partners (than individuals in the MONOG-min EDSA class), and more likely to have recently used GSNs and engaged in heavy drinking, suggesting that the second monogamy class might also represent slightly dysregulated individuals in monogamous relationships.

Consensual Nonmonogamous (CNM) Relationships Can Be Healthy

Consistent with the majority of previous research (see Barker & Langdridge, 2010; Rubel & Bogaert, 2015 for reviews), we found no significant differences in relationship functioning between monogamous and Open-CNM relationships. The latent group of Open-CNM relationships reported high levels of relationship functioning comparable to those of the two monogamous groups. This suggests that Open-CNM relationships that involve mutual consent, comfort, and communication surrounding EDSA can be healthy, satisfying relationships. Although the partners in these relationships have low interest in monogamy, are highly embracing of casual sex, are actively seeking new sexual partners (via GSNs), and are actively engaging in EDSA, the results suggest that they are doing this in a manner that maintains the quality and integrity of their primary relationships. In fact, as the Open-CNM were some of the longer relationships in the study yet continued to report high levels of relationship functioning, the results could suggest that EDSA conducted within truly Open-CNM relationship structures could vitalize nonmonogamous relationships, offsetting the natural decay in quality observed in meta-analyses of traditional marriage relationships (Karney & Bradbury, 1995; Mitnick, Heyman, & Smith Slep, 2009). Alternatively, the results could suggest that it is individuals in thriving relationships who are most able to successfully navigate the dynamics of CNM relationships with active communication and sensitivity. These results stand in stark contrast to a growing body of cross-sectional research linking EDSA to lower relationship quality in monogamous relationships (e.g., Atkins, Baucom, & Jacobson, 2001; Mark et al., 2011; Pereira et al., 2014; Whisman et al., 2007). Similarly, longitudinal studies have shown that lower relationship satisfaction and lower commitment significantly predict future EDSA (Drigotas, Safstrom, & Gentilia, 1999; Maddox Shaw, Rhoades, Allen, Stanley, & Markman, 2013). Taken together, this body of work begins to suggest that EDSA might serve as a reaction to relationship dissatisfaction and discord. However, the majority of studies examining EDSA have been conducted in samples of traditional relationships (e.g., married couples) where monogamy was inferred but not directly measured. By distinguishing a latent class of Open-CNM relationships, the current study provided critical insights as to the key factors which differentiate EDSA that could be potentially damaging to relationships or a sign of distress from EDSA that might be invigorating to relationships or a sign of relationship health.

Imbalanced EDSA Could Be a Sign of Relationship Discord

A majority of individuals in the group with primarily one-sided EDSA reported recent EDSA (61%) despite knowing that this behavior would likely upset their partners. This suggested that the EDSA reported by these individuals could potentially represent a form of infidelity as it might reflect a betrayal or at least a disregard for the comfort of their romantic partners. Consistent with this, the individuals in this group reported some of the lowest relationship and individual functioning. These findings align with previous research on infidelity (Atkins et al., 2001; Mark et al., 2011; Pereira et al., 2014; Whisman et al., 2007) as well as previous studies which found that monogamous couples with one partner engaging in EDSA had lower relationship quality when compared to strictly monogamous and CNM relationships (LaSala, 2004; Rodrigues et al., 2016). Taken together, this suggests that EDSA occurring in the absence of mutual consent, comfort, and communication could have devastating effects on relationship quality and/or could represent a maladaptive reaction to a failing relationship. However, the current findings on one-sided EDSA also highlight the importance of mutual comfort with any relationship structure. Thus, even in relationships with explicit and mutually agreed-upon nonmonogamous structures, a discrepancy in comfort with EDSA (and/or desire for monogamy) between the partners that results in only one partner engaging in EDSA could potentially be harmful to the ongoing quality of that relationship (as it could represent something of a callous disregard for the feelings and wishes of a romantic partner). Although we are aware of no research in this area to date, future studies should examine possible forms of such one-sided nonmonogamy occurring within samples of open relationships, swingers, and even polyamory to help uncover the correlates of discrepant relationship structures.

Nonmonogamy, Sexual Orientations, and Attachment Avoidance

Although the types of individuals drawn to various relationship structures are described in the Results section, a couple of findings bear further discussion. The CNM-open group had higher proportions of heteroflexible individuals (32.03%) and bisexual individuals (28.13%) than either of the two monogamous groups. This supports recent work that found that individuals in nonmonogamous relationships were more likely to identify as non-heterosexual (Balzarini et al., 2018a), suggesting that individuals identifying with non-heterosexual orientations may also be more embracing of nonmonogamous relationship structures.

The current results also suggested that individuals in the two maladaptive nonmonogamous classes were significantly higher in attachment avoidance than the other (more adaptive) relationship structures. These findings build on previous work which showed a greater willingness on behalf of individuals with high levels of attachment avoidance to engage in CNM, although that greater willingness did not necessarily translate into higher rates of actual participation in nonmonogamous relationships (Moors, Conley, Edelstein, & Chopik, 2015). The current results suggest that individuals with high levels of attachment avoidance might indeed seek out nonmonogamous relationships, potentially finding them more comfortable than traditional monogamy. In fact, given the lower rates of relationship functioning seen in those two relationship structures, the current results suggest that individuals with high levels of attachment avoidance might turn to EDSA (potentially without actively and openly discussing it with their partner) as a reaction to (or a form of coping with) relationship conflict or distress. Alternatively, these findings might suggest that individuals with high levels of attachment avoidance might have difficulty communicating about their EDSA, even when in nonmonogamous relationship structures, thereby eroding their relationship quality over time as their secrecy around that behavior builds.

Nonmonogamy and STI Risk

There were significantly higher lifetime rates of STIs in the individuals in relationships with nonmonogamous structures – showing elevated levels for all three nonmonogamous groups. Previous work in this area has shown mixed findings: Lehmiller (2015) found that individuals in CNM relationships showed comparable rates of STIs to those in monogamous relationships while Platteau et al. (2017) identified higher STI rates in swinging relationships (a particular form of open CNM). While the current results could potentially suggest that nonmonogamous relationship structures of any type might place couples at heightened risk for STIs, the larger correlates help contextualize these findings. The individuals in the nonmonogamous groups reported greater sexual sensation seeking and active partner seeking. Consistent with previous work (e.g., Balzarini et al., 2018b), they also had more embracing attitudes toward casual sex. Those findings suggest an alternative explanation for the higher rates of lifetime STIs observed, as those individuals might have engaged in more adventurous (and potentially unprotected) sexual activity with a larger number of partners prior to their current relationships. Looking at ongoing sexual risk behavior, the two nonmonogamy groups marked by lower comfort, consent, and communication showed significantly higher rates of condomless sex with new partners than the monogamous groups, while the open CNM group showed rates somewhere in the middle. This is consistent with work showing greater use of safer-sex practices in CNM relationships than in infidelity (Conley et al., 2012). Future work will be needed to determine if nonmonogamous relationship structures (hopefully defined with the COMMIT-4 within the Triple-C framework) might place partners at a greater risk for STIs contracted during that relationship. This alternative possibility (that the higher rates of STIs might have predated the current relationship) is consistent with findings that individuals in CNM relationship are more careful with their sexual health than those in monogamous relationships who engage in EDSA (e.g., Conley et al., 2012).

The Triple-C Model Will Likely Inform Work on Specific Forms of Nonmonogamy

As detailed in the introduction, the current study took a conceptually broad approach to identifying fundamental classes of relationship structure, grounded within the Triple-C model. As a consequence, we did not ask detailed questions about the agreed upon contexts in which EDSA could occur within nonmonogamous relationships. Consequently, our LPA analyses were able to identify more fundamental classes of adaptive and maladaptive relationship structures that would underlie all forms of monogamous and nonmonogamous relationships, providing support for the multidimensional conceptualization of structures based on the Triple-C model. Given that conceptual focus (and by design), our analyses would not have been able to discern differences between various forms of nonmonogamy like swinging, polyamory, and open/CNM relationships. Future work will be needed to determine the precise links between the fundamental forms of relationship structure presented here and those more specific forms of nonmonogamy. However, we conceptualize the Triple-C model as foundational to establishing a healthy and rewarding relationship structure. Thus, we believe that the classes that emerged in our analyses likely exist within any population of swinging, polyamory, or open/CNM relationships. In fact, we would assert that many of the non-significant or mixed findings in the previous literature are likely due to the fact that previous samples of open/CNM relationships (or swinging, or polyamorous relationships) had a mixture of the three classes of nonmonogamy identified here. In fact, couples engaging in occasional or limited swinging might actually view themselves as primarily monogamous and could end up being most appropriately classified in the Monogamous with low rates of EDSA class. Thus, without using the Triple-C model and the COMMIT-4 to disentangle the various forms of monogamy and nonmonogamy (even within a sample engaging in a very specific form of nonmonogamy, like swingers or polyamorous relationships), meaningful differences on relationship and individual functioning would likely be obscured. Thus, we would strongly encourage researchers studying one of these more specific forms of nonmonogamy to assess the components of the Triple-C model (or use the COMMIT-4) to augment their single-item assessment and classification of relationship structures prior to exploring how individual and relationship functioning might differ between monogamous and nonmonogamous relationships.

Strengths and Limitations

The current study provides a new conceptual framework to understanding relationship structures. The Triple-C model provides three crucial processes (mutual consent, comfort and communication) to help differentiate adaptive from maladaptive relationship structures, whether monogamous or nonmonogamous. Using the components of the Triple-C model in the first data-driven, person-centered analyses within a large and diverse online sample to be conducted in this literature, the current study uncovered five fundamental classes of relationship structures, likely underlying common self-identified relationship structures (i.e., monogamous, open relationships, swingers, polyamory). These fundamental relationship structures not only demonstrated marked differences in individual and relationship functioning, but also moderated the meaning of online partner-seeking (i.e., GSN app use), highlighting the salience of the Triple-C model.

Despite these strengths, the interpretation of the results is limited by a number of factors. First, the LPAs conducted were entirely exploratory. Thus, a critical first step in extending this work will be to replicate these relationship structure classes in a new dataset. Second, data were collected from only one partner in each relationship. This forced the analyses to rely on just one individual’s perceptions from each couple which may or may not align with the perspectives of their partners. Thus, to fully model relationship structures and directly assess their mutuality, future research should seek to cross-validate the latent classes identified within dyadic samples in which both romantic partners report on the relationship structure of the relationship (and on their relationship’s levels of functioning). Third, the survey was cross-sectional. Future studies could use longitudinal data to explore potential directions of causality. Fourth, the multivariate approach used to define the relationship structure classes created broader classes and did not permit the differentiation of CNM relationships into more specific categories, such as swinging and polyamory. Future studies could therefore include additional questions to allow for greater differentiation of the various types of CNM relationships. Fifth, although the current study was the first to introduce the Triple-C model and offered fairly compelling support for the relevance of that model, the study used a fairly brief and preliminary set of items to assess the components of the Triple-C model. As these analyses were exploratory and as most of the items were created specifically for this study, future studies could extend this work by evaluating the classes of relationship structure that emerge with a larger and more diverse set of items linked to the components of the Triple-C model. This would help to (1) confirm the relationship structures that emerged, (2) clarify links to specific forms of nonmonogamy (i.e., swinging, polyamory), and (3) identify the most effective items for classifying relationships. For example, future work could endeavor to determine if one or two-item assessments might be equally effective as the COMMIT4 at identifying these groups.

In addition to the limitations mentioned above, although the machine-learning derived COMMIT4 offers researchers a simple and brief method of fairly accurately (~93%) classifying participants into the five fundamental classes of relationship structures identified, meaningful heterogeneity might still exist within this parsimonious set of classes. Future researchers could therefore extend this work, for example, by examining how couples engaging in recent EDSA within one of those groups might differ from couples not engaging in recent EDSA within that same group. It should also be noted that, the sample was predominantly female (70.75%), white (78.65%), and college educated (55.42%) potentially limiting how well the results generalize. Fortunately, the sample contained 481 men and the LPA results suggested that those men were well represented across all of the groups examined. In fact, separate LPAs run in men alone yielded extremely similar classes of relationship, suggesting that the findings are comparably relevant for men. In addition, the sample contained 736 individuals with an associate’s degree or less and 343 non-White individuals, allowing for those groups to still be represented within the analyses. In addition, condomless sex was only assessed in a subsample of respondents who had recently reported meeting new individuals via social activity in an effort to minimize participant burden. This might have introduced biases and future research in this area should assess risky sexual behaviors in all respondents. Furthermore, the MSOCS items assessing sexual orientations did not provide clear options for individuals identifying as either asexual or pansexual. Although endorsement of those sexual orientations tends to be fairly low (Legate & Rogge, 2019), future work should assess sexual orientations in a more inclusive manner. Finally, given our focus of identifying the fundamental relationship contexts in which EDSA would be maladaptive vs. adaptive, the current paper focused primarily on sexual nonmonogamy. Future work will be needed to characterize emotional forms of nonmonogamy.

Conclusions

Despite these limitations, the current study highlights the diversity that exists within relationship structures, clarifying that while monogamy remains the prevalent structure in the United States, as much as 30% of relationships identify as nonmonogamous. The findings characterized three fundamental classes of nonmonogamous relationships likely underlying self-identified groups, thereby providing critical insights to help integrate a previous literature marked by largely mixed findings and heterogeneous operational definitions. We also delineated two distinct forms of monogamy, identifying a class of monogamous relationships at greater risk for recent infidelity. Finally, we developed a 4-item tool, the COMMIT4, to provide a short and straightforward tool for future researchers to classify romantic relationships into these latent classes. As the relationship groups displayed marked differences across the wide range of measures assessed, the COMMIT4 provides researchers with a simple method of incorporating diversity in relationship structures within their studies and their analyses.

Declaration of Conflicting Interests

The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

Open Practices

All IRB materials as well as the syntax files for the SPSS and Mplus analyses have been made publicly available via the Open Science Framework and can be accessed at https://osf.io/vx7rp/. De-identified data is available upon request to the authors via the same Open Science Framework.

Author Contributions

R. Rogge developed the study concept and collaborated with D. Crasta on the study design. R. Rogge was responsible for the IRB approval, implementation and data collection. All authors performed the data analyses and interpretation. F. Hangen drafted the manuscript and all authors provided critical revisions. All the authors approved the final version of the manuscript for submission and are responsible for its content.

Additional information

Funding

This work was supported by the NSF Graduate Fellowship [DGE-1419118];Office of Academic Affiliations Advanced Fellowship Program in Mental Illness Research and Treatment, Department of Veterans Affairs.