Self-Internet Infidelity and Partner Internet Infidelity. The Only Romanian Psychological Instruments Assessing the Online Extradyadic Behavior

The aim of this study was to validate the Romanian version of the Internet Infidelity Scale (IIS), and to determine its psychometric properties in Romanian couple relationships. Specifically, IIS assesses the severity of 44 specific acts on both the Self Infidelity and the Partner Infidelity versions. Methods: 675 participants completed the Romanian version of the Internet Infidelity Scale (RVIIS), the Infidelity Scale (DIS), the Extra-Dyadic Behaviors Inventory (EBI), and the Satisfaction with Life Scale (SWLS). We used the exploratory (EFA) and confirmatory factor analysis (CFA) to examine the structure and models fit for Self and Partner Internet Infidelity. The scale’s reliability, convergent and divergent validity were also tested. Results: The EFA and CFA analysis used for testing the matrix structure for Self and Partner Infidelity Scale revealed adequate models. An overall Cronbach’s alpha of .80 for Self-Infidelity Scale and .88 for Partner Infidelity Scale, showed good internal consistency. Positive levels of convergent validity for each scale revealed that the instruments measure what they have intended to measure. The good fit of the scales ensures a pronounced robustness of the construct. Conclusions: The validated Romanian version of the Internet Infidelity Scale (RVIIS) is a valid and reliable instrument in rating the severity of 6 specific behaviors on both the Self-Infidelity and the Partner-Infidelity


Introduction
Today, the Internet is clearly an essential component of our daily life, becoming one of the most exciting common places for social interaction and romance (Ben-Ze'ev, 2004). The advent of technological advancements and the magnetism of the virtual world allow people around the world to easily communicate.
Millions of such users may be married, seriously dating, or cohabiting individuals, and they may use the Internet to meet strangers, flirt, and many times engage in highly sexualized conversations (Mileham, 2007). The large number of couples involved in extra-dyadic Internet relationships and online infidelity is a growing problem (Gerson, 2011;Nooripour et al., 2016). The difficulty of such relationships has led couples to seek help from counselors and therapists. Computer-mediated relationships are as important and significant as direct face-toface romantic and intimate relationships (Atwood, 2005;Cravens et al., 2013).
An online relationship is generally defined as any sexual or romantic relationship initiated via virtual communication and continues through other electronic communication such as email, or Internet conversations that occur in online communities such as chat rooms, interactive games, and news groups (Hertlein & Piercy, 2012;Sharabi et al., 2021;Young et al., 2000). This online relationship becomes act of betrayal when an individual engages in secretive, romantic, or sexual behavior with a virtual partner, despite being engaged in a serious romantic relationship with someone in the "real" world (Norton et al., 2018;Young & de Abreu, 2011). The findings of Abbasi and Dibble (2021) suggest that partners should be cautious when making friends online and should take proactive steps to avoid the possibility of engaging in infidelity behaviors.
The technological sphere, including social media, cybersex, and written digital communication like email, text, and chat rooms, has been rapidly growing and it shows technology as an infidelity facilitator (Rothstein et al., 2021).
Theory of Planned Behavior (TPB) (Ajzen, 1985(Ajzen, , 1991) is a very powerful and predictive cognitive model for explaining online infidelity. It evolved from the Theory of Reasoned Action (Fishbein & Ajzen, 1975) which posited "intention to act" as the best-predictor of behavior. In addition to attitudes and subjective-norms, TPB adds the concept of perceived behavioral control, which originates from Self-Efficacy Theory (SET), proposed by Bandura in 1977. The premise of the TPB is that individuals make rational-decisions to engage in specific-behaviors based on their-own beliefs about the behaviors and their-expectation of a positive outcome after having engaged in the behaviors. TPB hypothesizes that infidelity happens because of the opportunity, as well as the intention to cheat. According to Ajzen (2002), an intention to perform a behavior is determined by three-components: 1) attitude toward a behavior (beliefs about a specific-behavior and its-consequences); 2) subjective-norm (normative-expectations of other-people who are important to the actor regarding the behavior), and 3) perceived-behavioral-control (the perceived-difficulty or ease of performing the behavior). Thus, the Internet infidelity takes on a special emphasis because partners are often subjective and self-serving and will take their needs into consideration rather than their partner's, thus justifying self-infidelity as more acceptable than partner-infidelity (Docan-Morgan & Docan, 2007). Previous research has defined betrayal as either emotional and sexual, or online. But Internet infidelity can also include sexual and emotional infidelity (Cooper et al., 2003;Henline et al., 2007;Liu et al., 2020;Suliakaite, 2009 In the next phase, the 65-item questionnaire was given to researchers in the field of internet relationships to examine its validity. As many items of the master list were eliminated due to repetition, the total number of scenarios decreased from the initial list. The final version of the scale had 44 items, and asked participants to rate the severity from 1 (not infidelity) to 5 (highest degree of infidelity).
Factor analysis led to the extraction of two factors after conducting a sample of 208 participants. The first factor was superficial/informal acts such as talking about everyday subjects, sports, news, telling jokes and sending friendly emails to the virtual partner. The second factor was involving/goal-directed activities such as online emotional or sexual relationships with a virtual partner, flirting, expressing excessive intimacy to the Internet partner, and sending photos and personal videos.
In the Internet Infidelity Scale (IIS), the first version asked respondents to imagine that they were committing the acts and to express their attitude toward each of these actions, whereas the second version asked subjects to imagine their partners committing any one of these secret behaviors. Self-internet infidelity refers to personal acts of infidelity and highlights the self's perspective (e.g., Purtarea unei conversaţii online despre viaţa dvs. sexuală, cu o persoană ȋntâlnită pe Internet/Having a conversation about your sex life in an Internet chat room with a person you met online), whereas partner internet infidelity refers to partner's acts of infidelity, also highlighting the self's approach (e.g., O relaţie intimă, pe care partenerul/partenera dvs. o are cu o persoană ȋntâlnită online/Your partner having an intimate relationship with a person he/she met on-line). Likewise, single participants were requested to complete the scale with an imaginary partner. The authors concluded that when the participants were rating goal driven or oriented acts, the ratings were more severe than if they were rating superficial or informal acts online (Connolly, 2018). On the other hand, Suliakaite (2009) showed that self-differentiation is associated with attitudes toward internet infidelity. Higher differentiated people considered less online behaviors to be unfaithful (Suliakaite, 2009), thus Lithuanians perceiving internet infidelity to be a multifaceted phenomenon. Nooripour et al. (2016) concluded that for most people having online extradyadic relationships, these are as real and as important as face-to-face-relationships. But because these findings are based on North-Central America and Northern Europe, it is an open question whether they can be generalized for Eastern Europe due to the differences in the contextual factors (socio-economic status, culture) (Rusu et al., 2015). Thus, the aim of this study was to adapt the Internet Infidelity Scale (IIS), determine its reliability and validity, and verify the adequacy of the Romanian adapted version of the six-dimensional scale.

Research Design
The study's procedure and the administered tools were fully compliant with the Declaration of Helsinki and with the University's Code of Ethics. To achieve the aim of this study, a cross-section correlational design type was used. Data collection was carried out by questionnaires through the main social networks as a method of recruitment, including on Facebook, Instagram, Twitter, Google+, and WhatsApp. This announcement directed participants to the online survey, where they read the informed consent, which reported the following: 1) the purpose of the study, 2) the inclusion criteria, 3) the voluntary nature of the study, 4) the possible risks, and 5) the right to withdraw from the study at any time. To protect the privacy of the participants, the questionnaires were completed anonymously (participants did not have to provide their name, phone number, or email).

Participants
A total of 675 participants (186 males, 489 females) with ages ranging from 23 -73 (mean = 40.94; SD = 11.09) completed an online web-based survey. Some of the participants were recruited from different social media networks and some others from psychology and communication classes from two Romanian universities. Using the Internet to conduct quantitative research provided us the opportunity to transfer survey responses directly into an Excel database, eliminating transcription errors and preventing survey alteration by the survey respondent. Socio-demographic data included age, education level, marital status, relationship length. Sampling was based on convenience. Both versions were completed by participants after they were ensured that their participation in the study was anonymous and confidential. The purpose of the research was explained to every participant, and the data were completed on a voluntary basis. To participate in this study, we established the following inclusion criteria: being a resident of Romania, being 18 years of age or older, and having a heterosexual committed relationship (married or cohabitation).

The Internet Infidelity Scale (IIS)
The English version of the Internet Infidelity Scale (IIS) was translated into Romanian, and its validation process was performed through the forward-backward translation strategy and following the guidelines provided by the literature (Beaton et al., 2000;Dunn et al., 1994). The scale was developed by Docan-Morgan and Docan (2007) to measure which types of acts involving the Internet are considered most severe, sex differences in the perception of infidelity, and the evaluation of infidelity when one commits it versus one's partner. The Romanian version of the scale was derived from the English version and was validated by the authors. The items were formulated using simple and appropriate language relative to the concept. However, according to previous literature on internet relationships, chat rooms are only one of the means for users to meet each other online, hence the term "chat room" in the Internet Infidelity Scale was changed to "online". For example, "having cybersex with a person you met in a chat room" was changed as follows: "having cybersex with a person you met online". Also, the ninth item: "having friendly conversations in a chat room titled 'Married and Looking'" was modified when translated into Romanian, as follows: "having friendly conversations online", due to the Romanian culture and reality.
Two 44-item versions of the scale were developed, the Self Infidelity (IISI) and the Partner Infidelity (IISP) Questionnaires. The first version asked participants to imagine that they were committing the acts, whereas the second version asked them to imagine that their partner was committing the behaviors. When rating the items, participants were asked to assume that they/their partner did not know that their partner/they were engaging in those behaviors. Respondents who did not have a partner were asked to rate the items as if they had a partner. Participants were also told to assume that the person they/their partner were interacting with online was someone to whom they/their partner could be romantically attracted (Docan-Morgan & Docan, 2007). The severity of the rank scenario rated from 1 = not infidelity; 2 = slight degree; 3 = considerable; 4 = strong degree; 5 = highest degree. A score of 1 -2 will indicate a low score of internet infidelity, a score of 3 is neutral, and a score of 4 -5 will indicate a high score of internet infidelity.
The first factor (15 items)-superficial/informal acts-accounted for 26.44% of the variance and the alpha reliability of these items was α = .95. The second factor (12 items)-involving/goal-directed acts-accounted for 23.67% and its alpha reliability was α = .92. For the present sample, the Cronbach's alpha for the IISI was .98 (mean = 2.30; SD = .90) and .99 (mean = 2.51; SD = 1.05) for the IISP.

The Infidelity Scale (DIS, Drigotas et al., 1999)
The Infidelity Scale, developed by Drigotas et al. (1999), consists of 11 items. Participants were asked to evaluate the level of attraction to a person, other than their partner, in either a current or past relationship on a 9-point Likert scale (0-not at all attractive; 8-extremely attractive). The items assessed both emotional and physical intimacy. In addition to using the emotional and physical intimacy scores, previous research by Drigotas et al., (1999) performed factor analysis, which revealed that nine of the 11 items loaded highest on a single factor that they labeled "composite infidelity". The 2 items that did not seem to relate to this factor were questions 2 and 6, which measured the speculation of how much the other person was attracted to the participant and initiation of the mutual attraction. However, because the other 9 items loaded so strongly on composite infidelity (factor loadings ranging from .61 to .85) the 9 items were averaged to form a composite infidelity score, with the maximum score being

Extra-Dyadic Behaviors Inventory (EBI; Luo et al., 2010)
Extra-dyadic Behaviors Inventory (EBI; Luo et al., 2010) consists of 23 items assessing face-to-face EDB (e.g., "kissing"; "romantic date"; "received oral sex"; "vaginal intercourse") and 13 items assessing online EDB (including internet and phone interactions, e.g., "spent time online with romantic interest"; "shared sexually provocative pictures"; "phone sex"). Participants were asked to report how often they had engaged in each of the listed behaviors while in their current relationship with someone (of the opposite sex) other than their current partner. Have participated in this behavior more than once with the same person; and 5 = Have participated in this behavior with different people". In this study, we used the two-factor structure of the online EBI, namely: online sexual EBI and online emotional EBI. Cronbach's alpha was .92 (mean = 1.97; SD = 1.01).

Satisfaction with Life Scale (SWLS; Diener et al., 1985)
A 7-point Likert style response scale (ranging from 1 = strongly disagree to 7 = strongly agree) was used to afford respondents an array of response options. The possible range of scores is therefore 5 to 35, with a score of 20 representing the neutral point on the scale. Scores between 5 and 9 indicate that the respondent is extremely dissatisfied with life, whereas scores ranging between 31 and 35 indicate that the respondent is extremely satisfied with life. Scores between 21 and 25 represent slightly satisfied, and scores from 15 to 19 are interpreted as falling in the slightly dissatisfied range. For the present sample, the Cronbach's alpha for SWLS was .92 (mean = 26.30; SD = 6.35).

Statistical Analysis
We used exploratory (EFA) and confirmatory factor analysis (CFA) with SPSS v. 20 to examine the matrix structure and the fit of the IISI and IISP. We chose 1) the absolute match measures (CMIN/DF) that determined the degree to which the model predicted the observed correlation matrix and whose value was recommended to be below 5, as well as the RMSEA that indicated approximate fits of the pattern in population. We also chose 2) the incremental measures (TLI, CFI) that compared the proposed model to a baseline model that all other models should overtake and that indicated the discrepancy between the two models (Huza, 2019;Nokelainen, 2009 Table 1 depicts the socio-demographic characteristics of the participants in the study.

Exploratory Analysis
The first step was to test the structure matrix of the two factors proposed by the authors of the 44 items scale. The EFA revealed appropriate indicators: the adequacy measure of sampling KMO (.829), Bartlett's test of sphericity (χ 2 = 26168; p = .00). The two components accounted 68.83% of the variance. Testing the fit of model with CFA, we found that this model is inappropriate for Romanian people (TLI = .55; CFI = .59; RMSEA = .222). The second step was to find a new matrix structure for Self and Partner Scale. We ran the EFA analysis, and we found appropriate models. The new models keep the two factors (Table 2 and  Table 3).

Condition for EFA Running for the Self and Partner Scales
The KMO test (.746) and Bartlett's test of sphericity (χ 2 = 1926. 2; p = .00), revealed appropriate indicators for running the analysis. A rotated analysis with varimax rotation was conducted with two components, accounting for 76.00% of variance. Each factor was composed of three items and accounted 52.51% of the variance (the first) and 23.49% (the second).
The KMO test (.756) and Bartlett's test of sphericity (χ 2 = 4806.94; p = .00), revealed appropriate indicators for running the analysis. A rotated analysis with varimax rotation was conducted with two components, accounting for 90.34% of variance. Each factor was composed of three items and accounted 63.27% of the variance (the first) and 27.07% (the second).

Confirmatory Analysis of the Scale
For testing the fit of the Romanian Version of the Internet Infidelity Scale for Self and Partner, we conducted a confirmatory analysis with Amos 20.0. The results revealed a good fit of the structural models. Table 4 depicts the CFA results analysis. The models are shown in Figure 1. In Figure 1, RVIISS represents the theoretical model for Self-Scale and RVIISP represents the theoretical model for Partner Scale. Both models have two factors: superficial/informal acts and involving/goal-directed acts. The RVIISS scale structure is revealed in Table 2 and the RVIISP scale structure is revealed in Table 3.

Reliability and Validity
The reliability analysis revealed a good Cronbach's alpha of .80 for the Self-Scale. The corrected item to total correlation was greater than .3 (see Table 5). Both factors have high reliability values: .86 for the first factor (superficial/informal acts, SIA) and .82 for the second factor (involving/goal-directed acts, SDA). The reliability analysis revealed a good Cronbach's alpha of .88 for the Partner Scale. The corrected item to total correlation was greater than .3 (see Table 6). Both factors have high reliability values: .91 for the first factor (superficial/ informal acts, PIA) and .97 for the second factor (involving/goal-directed acts, PDA).
Convergent validity for Self-Scale was supported by a significant correlation (.47**) between the RVIISS total score and the DIS total score, respectively (.43**) between the RVIISS total score and the EBI total score. Convergent validity is shown in Table 7. Divergent validity was supported by a significant correlation between the RVIISS total score and SWLS total score. Divergent validity is shown in Table 8.
Convergent validity for the Partner Scale was supported by a significant correlation (.46 **) between the RVIISP total score and the DIS total score, respectively (.42 **) between the RVIISP total score and the EBI total score. Convergent validity is shown in Table 9.
Divergent validity was supported by a significant correlation between the RVIISP total score and SWLS total score. Divergent validity is shown in Table  10.   (Your partner joking around on Instant Messenger with a person he/she met on-line). .

.885
Total Alpha = .88. Table 7. Convergent validity for Self-Scale.  Note: RVIISS-total score of the Romanian Version of the Internet Infidelity Self Scale; SIA-total score of the superficial/informal acts; SDA-total score of the involving/goal-directed acts; SWLS-The Satisfaction with Life Scale; **p < .000.   Our findings confirm Docan-Morgan & Docan's (2007) results that assume that informal/superficial acts are considered less severe than involving/goal-directed acts. We tested this assumption by Wilcoxon test for Self and Partner Scale.
The second assumption assumes that partner's infidelity will be seen as more severe than self-infidelity will. Our results confirm this assumption: z = −17.93, p = .00, Mean rank RVIISS = 103.37, Mean rank RVIISP = 311.72. For all analyses, the highest values (sum of ranks) correspond to the positive ranks, which means that the ranks of the variables: SDA, PDA, RVIISP are higher than the ranks of the variables: SIA, PIA, RVIISS. Table 11 depicts how men and women perceive self-infidelity and partner's infidelity.
For all the analyses, the highest values (sum of ranks) correspond to the positive ranks, which means that the ranks of the variables: PIA, PDA, are higher than the ranks of the variables: SIA, SDA.
To find out whether the two scales discriminate according to age and relationship length, we divided the participants into age categories and different relationship lengths. Table 12 and Table 13  Note: SIA-total score of SIA; PIA-total score of PIA; SDA-total score of SDA; PDA-total score of PDA; **p < .00; *p < .05.  For all the analyses the highest values (sum of ranks) correspond to the positive ranks, which means that the ranks of the variables: PIA, PDA, are higher than the ranks of the variables: SIA, SDA. Docan-Morgan & Docan (2007) proposed the hypothesis that women will rate acts of Internet infidelity as more severe than men will rate acts of Internet infidelity. Our study confirms the hypothesis for men and women with age less than 30 years, but only for self-involving/goal-directed acts (U = 1433.00; z = −2.09, p = .037; Mean rank MEN = 59.65; Mean rank Women = 76.47). The severity for each item of the Romanian version scales is shown in Table 14.  The EFA of the Self and Partner Scales show appropriate models. The 6 items of Self Scale (RVIISS) show a good Cronbach's alpha of .80. Both its factors have high reliability values: .86 for the first factor (superficial/informal acts, SIA) and .82 for the second factor (involving/goal-directed acts, SDA). The 6 items of Partner Scale (RVIISP) show also a good Cronbach's alpha of .88. Both its factors have high reliability values: .91 for the first factor (superficial/informal acts, PIA) and .97 for the second factor (involving/goal-directed acts, PDA). A significant correlation between the RVIIS for Self and Partner with DIS and EBI has indicated a good convergent validity. (.47**, respectively .43** for Self-Scale and .46**, respectively .42** for Partner Scale). From the items of the original scales, the Romanian version of the two scales preserves only those items (6) that ensure a pronounced robustness of the construct. The CFA confirms this robustness through the high values of indicators: RMSEA = .067 [.043; .091]; TLI = .977; CFI = .988 for Self-Scale and RMSEA = .072 [.047; .098]; TLI = .988, CFI = .995 for Partner Scale. It shows that the models are appropriate and do not falsify the reality. The high psychometric qualities of the Romanian version of the Internet Infidelity Scale (RVIIS) for Self and Partner recommend it for both screening and research. The two versions of the Internet Infidelity Scale (the Self-Scale and the Partner Scale), along with the behavior's description and the rank scenario, are depicted in Appendix.

Discussions
The aim of the present study was to test the psychometric properties of the Ro-  (2007) found a good reliability for the Internet Infidelity Questionnaire (.92 -.95 for subscales). Our study showed good Cronbach's alpha of .8 (.86 -.82 for subscales) for the Self Scale, and .88 (.91 -.97 for subscales) for the Partner Scale. The Romanian Version of the Internet Infidelity Scale for Self and for Partner maintains the two factors of the original version. If the original version keeps the same items for self and partner (e.g., Joking around on Instant Messenger with a person you met on-line; Your partner joking around on Instant Messenger with a person he/she met on-line), the Romanian version contains both common and specific items (see Table 5 and Table  6). It shows us that we judge our partner's infidelity in qualitative terms as well (e.g., Having a conversation about your sex life in an Internet chat room with a person you met on-line versus Your partner having an intimate relationship with a person he/she met on-line). The Romanian version of the Internet Infidelity Scale is related to gender, marital status, age categories and relationship length, and it discriminates between informal/superficial acts and involving/ goal-directed acts. As a rule, both men and women regardless of marital status, evaluate the partner's infidelity as more severe than their own (both superficial/informal acts and the involving/goal-directed acts). Our results agree with the findings of other authors (Abdi et al., 2012;Docan-Morgan & Docan, 2007;Shrout & Weigel, 2020;Vossler & Moller, 2020). Age can be a factor in influencing the judgment of infidelity. Our findings show that men and women over 50 years old have similar judgments about infidelity only in terms of superficial/informal acts. Regarding the relationship length, only men under the age of 30 and those over the age of 50 evaluate their own informal/superficial acts and those of their partner's with the same severity. These results are in accordance with Whitty (2003) that highlights age differences and relationship status in attitudes towards sexual infidelity. The hypothesis that younger people rate acts of internet infidelity as more severe than older individuals is in accordance with Whitty (2003) that assumes that younger people were more likely to rate sexual acts as acts of betrayal than older individuals. Our scale indicates that women evaluate infidelity more severe than men. That is in accordance with Moreno, and Kahumoku-Fessler's (2018) results that show that females rated sexual and emotional behaviors higher than males.

Strengths and Limitations
There are strengths in the design and methodological approach of our study (e.g., the sample size, EFA, CFA, divergent and convergent validity) which increase our confidence in the results. Nevertheless, the current study comprises several limitations including the use of self-report measures, the lack of random sampling, and of test-retest reliability. The convenience sample used in this study that does not use the random sampling suggested that future research should analyze the factor structure of questionnaires in populations with different characteristics. Also, the study had low sample variability since subjects belonged to six specific geographical areas of the country. To validate the predictive capacity of the scale, it is necessary for future studies to analyze scale predictability with respect to other factors (e.g., intention and behavior; pornography consumption and online sexual activities). Lastly, future research should include longitudinal research, evaluate the test-retest reliability and predictive validity of the IIS.

Conclusion
The Romanian version of the Internet Infidelity Scale for Self and Partner showed a high reliability and a good convergent validity. The good fit of the two scales recommends its use in detecting internet infidelity. The high psychometric qualities of the Romanian Version of RVIIS Scales recommend it for various studies. To confirm the robust psychometric characteristics of the RVIIS, future studies should include homogenous populations as well (e.g., only adolescents, only adults, only married people). The RVIIS scales can detect the online extradyadic behaviors and help many people seek specialized help to reconsider and strengthen the couple and family relationships.