Psychometric properties of the Multidimensional Jealousy Scale ( MJS ) on a Serbian sample

The goal of the current study was to examine psychometric properties of the Serbian version of the Multidimensional Jealousy Scale (MJS, Pfeiffer & Wong, 1989). For this purpose, 500 participants, aged 18–40, were asked to complete this version of MJS. In order to examine external validity of MJS, participants were asked to complete The Rosenberg’s self-esteem scale, The Big Five Inventory and the Love Attitudes Scale. Confirmatory factor analysis supported the three-factor structure of the instrument, but required the addition of a number of error covariance arguments to achieve fit. Another model was tested, after which acceptable fit levels were achieved with smaller number of error covariance arguments in the model. Internal consistency of subscales was high. Correlations with neuroticism, self-esteem and mania love styles are consistent with the theoretical expectations and previous results, suggesting good external validity of the instrument.


Highlights
• Measurement properties of the Serbian version of MJS were studied.
• Confirmatory factor analysis supported the three factor structure of the MJS.
• Minor changes to model to achieve fit with less error terms.
• Correlations with three external variables support external validity of MJS.
The phenomenon of jealousy has been described for centuries and most of these descriptions treated it as a part of emotional passion, but no single definition has gained general acceptance.
Alternative definitions of jealousy as a psychological construct exist, some treating it as a one-dimensional construct (ex.Mathes & Severa, 1981;White, 1981b) and others treating it as a multidimensional one (ex.Pfeiffer & Wong, 1989).Although there are several questionnaires which are used to quantify the experienced jealousy (ex.Bringle, Roash, Andler, & Evenbeck, 1979, White, 1981b), these questionnaires are mostly one-dimensional and do not discuss qualitative aspects of jealousy.
The majority of current authors believe that multidimensional approach to jealousy leads to a better understanding of experiences of jealousy (Buunk & Dijkstra, 2006;Harris, 2009).However, despite the agreement on the multidimensional nature of romantic jealousy, there is no agreement on what dimensions jealousy consists of.Proposed factor structures vary significantly between approaches, studies and measurement instruments used (Gehl, 2010).In the case of the Multidimensional Jealousy Scale (Multidimensional Jealousy Scale, Pfeiffer & Wong, 1989) jealousy factors are based on the theoretical model of jealousy proposed by White and Mullen (White & Mullen, 1989).
White and Mullen (White & Mullen, 1989) define jealousy as a "complex of thoughts, emotions and actions that follows loss of or threat to self-esteem and/or the existence or quality of the romantic relationship" (p.9).These authors, following the Lazarus' appraisal theory of emotions (Lazarus, 1984), suggested that the experience of jealousy should include cognitive evaluation, emotional reaction and behavioral manifestation.According to the author's point of view, faced with the situation that is a trigger of jealousy, an individual makes a cognitive evaluation of the situation and determines whether there is a potential rival relationship as well as the level of threat the rival constitutes (White & Mullen, 1989).Cognitive jealousy means worrying about fidelity of the partner or his commitment to the romantic relationship (Pfeiffer & Wong, 1989).Cognitive evaluation is further accompanied by different emotional reactions.Emotional jealousy is defined as a level of anxiety that the person, who is exposed to situations that cause jealousy, suffers (White & Mullen, 1989).While cognitive jealousy represents lack of trust in the partner's commitment, emotional jealousy includes affective reaction to that kind of situation (Guerrero, Eloy, Jorgensen, & Andersen, 1993).However, both cognitive and emotional jealousy present internal experience of an individual in contrast to the behavioral aspect of jealousy that actually refers to the way in which people express jealousy with their acts and behavior (Guerrero & Andersen, 1998).
Behavioral jealousy refers to the expression of jealousy on a behavioral plan.In addition to that, expressing jealousy can be manifested through communication and other types of behavior that may include the partner (White & Mullen, 1989), but need not, as for example in spying on the partner (Guerrero, Andersen, Jorgensen, & Eloy, 1995).
The efforts to operationalize the multidimensional model of jealousy (White & Mullen, 1989) have led to the construction of the Multidimensional Jealousy Scale (Pfeiffer & Wong, 1989).Although the authors of MJS relied on this model of jealousy during the construction of the scale, the model actually applied in the scale included certain changes.In cognitive domain, the authors included paranoid suspicions and worries about partner's fidelity.They emphasized that emotional reaction can occur as a reaction to a real but also an imaginary threat to the romantic relationship, i.e. as an emotionally conditioned response to certain situations.Finally, behavioral jealousy, primarily defined as cognitive and behavioral strategies used by the individual to cope with the emotional excitement, includes different protective and detective actions in the MJS model compared to the model proposed by White and Mullen (1989).These actions include interventions which have the goal of making sure no contact occurred between the partner and the rival or to prevent that type of contact (Pfeiffer & Wong, 1989).Another significant change presented by the authors of the scale refers to the belief that the three dimensions of romantic jealousy are in interaction and that they occur simultaneously, before developing into a sequential process starting with suspicious thoughts through emotions through action, as it was primarily proposed by White (White, 1981a).
In a series of studies, Pfeiffer and Wong (Pfeiffer & Wong, 1989) have developed MJS as a measure of romantic jealousy on a Canadian sample.MJS includes 24 statements, eight for each subscale, in accordance with the theoretical model.After the scale construction, a series of exploratory factor analyses was conducted (EFA) and a stable, three-factor solution that corresponds to the cognitive, emotional and behavioral jealousy has been proposed (Pfeiffer & Wong, 1989).
The goal of the current study was to explore psychometric properties of the Serbian version of MJS created by authors of this paper, on a sample from Serbia and study both internal and external validity of the instrument in this population.

Participants and procedure
The sample consisted of 500 participants, aged 18 -40 (94.8 % of the sample belonged to the age group from 25 -35 years).Majority of the subjects, 395 of them, were of female gender.71% reported being in a relationship or marriage (57.6% and 13.4%, respectively), whereas, the rest, 29% of the sample, reported not having a partner.Participants that reported having no partner were asked to give their answers based on their experiences from previous romantic relationships.
The procedure of data collection was online sampling, using the snowball sampling procedure, where existing participants asked their acquaintances to also become participants in the study.

Instruments
For measuring jealousy the Serbian version of the Multidimensional Jealousy Scale -MJS (Multidimensional Jealousy Scale, Pfeiffer & Wong, 1989) was used.MJS is a 24item scale, where there are 8 items in each of the three subscales aimed to evaluate three components of jealousy: cognitive, emotional and behavioral.The cognitive scale evaluates to which extent the individual has worries and doubts regarding partner's fidelity.For the cognitive subscale, participants indicated how often certain thoughts about their partner occurred, with responses ranging from 1 (never) to 7 (all the time).A sample item is, "I suspect that my partner may be attracted to someone else."The emotional subscale measures the strength of the jealous emotions in situations that cause the experience of jealousy, such as: "My partner hugs and kisses someone of the opposite sex.".There are 8 items that measure the level of emotional anxiety, where the participant evaluates from 1 (very pleased) to 7 (very upset) his or her emotional experience in certain situations.The subscale related to the behavioral aspect of jealousy measures the frequency of actions and activities that represent expressions of jealousy, such as looking through partner's pockets, checking and questioning others about partner's movement etc..The participant reports how often he or she is involved in that type of actions on a scale from 1 (never) to 7 (all the time).The authors of the scale claim good reliability of the scale and of all subscales specifically: the cognitive, emotional and behavioral (Cronbach's alpha .92,.85,.89,respectively) (Pfeiffer & Wong, 1989).
Beside jealousy, for purposes of examining external validity of MJS, dimensions of personality, self-esteem of the examinees and love styles were also assessed.
The Big Five Inventory -BFI (Big Five Inventory; John et al., 1991, according to John & Srivastava, 1999) consists of 44 statements where the participant expresses his or her level of agreement on a 5-point Likert-type scale.The questionnaire is used for evaluation of the Big Five personality dimensions: neuroticism, extraversion, openness to experience, agreeableness and conscientiousness.Authors of BFI report high internal consistency of the scale (Cronbach's alpha = .84;α = .88;α =.81; α = .79;α = .82,for each dimension respectively).(John & Srivastava, 1999).
The Rosenberg Self-Esteem Scale -RSE (The Rosenberg Self-Esteem Scale, Rosenberg, 1965) was used for obtaining a measure of participants' self-esteem.RSE is a one-dimensional scale and measures global self-esteem.The scale consists of 10 statements with the instruction to participants to report on a 5-point Likert-type scale on the extent in which the statement refers to them.The scale has excellent test-retest reliability over a period of 2 weeks with correlations of .85 and .88(Rosenberg, 1979) Love Attitudes Scale -LAS (Love Attitudes Scale, Hendrick & Hendrick, 1990) was used for evaluation of the love styles.The instrument consists of 42 statements and the participants uses a 5-point Likert-type scale to express degree of agreement with each of them.The results indicate the level of expression of the following love styles: Eros (passionate love), Ludus (love as a game), Storge (friendly love), Pragma (rational love), Mania (possessive love) and Agape (unselfish love).Authors of the scale report alpha coefficients for the love style scales to range from .74 to .84 (Hendrick & Hendrick, 1990).

Results and discussion
First, internal consistency and descriptive statistics measures were calculated for each of the scales used.These results are presented in table 1.Also, item-wise descriptive statistics were calculated for each of the MJS items and these are presented in table 3. Results show that all the scales used in the study have adequate internal consistency on the current sample.The only exceptions are measures of Ludus and Storge love styles, internal consistency of which is somewhat lower than that of the other scales.Relative to MJS scales, obtained internal consistency measures are both adequate and similar to those obtained in previous studies using different language versions of MJS.Pfeifer and Wong (1989) reported Cronbach's alpha coefficients of .92., .85 and .85 for cognitive, emotional and behavioral jealousy scales respectively, while Elphinston et al. (2011) reported alphas of .77,.81 and .70respectively on their sample.Additionally, Lord-Kaiser-Caffrey coefficients of reliability of the first main component were calculated for the jealousy measures and these were .902,.836and .831for Cognitive, Behavioral and Emotional jealousy, respectively.
Relative to item-wise results of internal consistency analysis (table 3), these show that there is no single item or group of items the removal of which would result in a substantial increase in internal consistency reliability of any of the three scales.
It should also be noted that distributions of all three variables are leptokurtic with Cognitive and Behavioral Jealousy score distributions being extremely so.This was due to participants giving minimum score answers (i.e.not agreeing with) on most of the items on these scales.This response scheme resulted in variability of answers being restricted and thus can be reasonably expected that the covariance based statistics, like internal consistency measures for example, would have been even higher had this variability restriction not taken place.
Next, latent structure of MJS on the current sample was examined.A factor model of relations between item responses and jealousy constructs created for the procedure of confirmatory factor analysis was created and tested.The model specified the existence of three first order factors influencing eight items each, and that the factors were correlated with each other.Thus, model represented the theoretical structure of the MJS, the three subscales and their covariances replacing the overall score.Results of testing the fit of this model to the data indicated a relatively poor fit.Examination of the residual covariance matrix showed substantial covariance between items not accounted for by the model, and these covariances were added to the model as error covariance terms.The model thus created is presented in figure 1.  Fit indices for the two models are presented in table 2.  Δα -change in the Cronbach's alpha value of the scale the item belongs to if that item is removed from the scale.Factor loadings lower than .3are not listed in the table.Model 1 -standardized regression weights for factor-item relations in the first model.Model 2 -standardized regression weights for factor-item relations in the second model.Model 2 additions -standardized regression weights for relation between the emotional jealosy factor and the two items that were added to that factor in the second models Relative to factor structure of the MJS, results generally support the internal validity of the MJS.Although a number of inter-item error covariances were obtained, the factor model with three distinct jealousy subdimensions holds.The most significant deviation from the theoretical model is the behavior of items 2 and 6, which, apart from being indicators of cognitive jealousy, could be construed as being indicators of emotional jealousy as well, although with lower factor loadings judging by the standardized regression weights.Inspecting the contents of these two items, it can be noticed that these two items require the examinee to assess his feelings (if they are worried) about a certain aspect of behavior of other people, unlike other indicators of cognitive jealousy that require the participant to specify beliefs or thoughts about the partner's behavior or behavior of a third person.It may be hypothesized why these items double as indicators of emotional jealousy factor as well.Since this assignment of two items to a different subscale also meant a different way total subscale scores are calculated, comparison between the original subscale scores and the standardized weighted composites calculated according to model 2 was conducted.Original composite scores, calculated with uniform weights and with items 2 and 6 only calculated towards the Cognitive Jealousy score were compared with the subscale scores calculated according to model 2 i.e. with weighted composites in which model 2 regression coefficients were used as weights on standardized item scores, and items 2 and 6 included in calculation of both Cognitive and Emotional Jealousy scores.The results showed that the two sets of scores are almost identical, with correlations between equivalent scores calculated with these two methods ranging between .986 and .995.This clearly showed that for all practical purposes, original MJS composite scores are as good as regression weighted scores calculated according to model 2. This supports the practical use of the original MJS score calculation method, in spite of the differences in proposed model presented here.
Due to results from numerous previous studies implying differences between genders in partner relations functioning and especially related to jealousy (e.g.Bendixen, Ottesen Kennair, & Buss, 2015;Edlund & Sagarin, 2017;Lemmers-Jansen, Krabbendam, Veltman, & Fett, 2017;Zandbergen & Brown, 2015), a question of whether models described hold for both genders naturally arose.To answer this question, both models were tested on both male and female subsamples.Results showed that fit indices remain at levels indicating acceptable fit only on the female subsample and that they become substantially lower when the male subsample is considered.This is the case for both Model 1 and Model 2. Upon examination of the regression weights of both Model 1 and Model 2 on the male subsample, it can be seen that all the factor-item relations remain statistically significant, but that some of the error covariance term coefficients become low and statistically nonsignificant.Also a certain amount of new inter-item residual covariance that is not unaccounted for by the model appears, but none of it high enough to justify inclusion in the model, neither as a new factor-item effect nor as an error covariance term.
In this situation it is important to point out that this lack of fit on the male subsample does not come from the changed relations between factors and items, but from changes in error covariances between items.Also, it should be noted though that part of the reason lies in the fact that the male subsample is 105 participants only, meaning that an effect has to be larger to become statistically significant than is the case on the female subsample of 395 or the whole sample.
External validity was tested in the current study by calculating correlations of Serbian MJS scores with a number of other external variables, correlations of which with jealousy constructs measured by other language versions of MJS have been reported in previous studies.Relative to external validity of the examined version of MJS on the current sample, obtained results are fully supportive.These correlations are presented in table 4. It can be seen from the table that the correlation between jealousy and neuroticism is positive and of moderate intensity, while the correlation between jealousy and self-esteem is negative and low, which is in accordance with previous results (Buunk, 1997;Gehl & Watson, 2003;Guerrero, Spitzberg, & Yoshimura, 2004, Khanchandani & Durham, 2009;Mullen & Martin, 1994;Xiaojun, 2002;White & Mullen, 1989).An explanation that is often given for the positive relation between neuroticism and jealousy and the inverse relation between self-esteem and jealousy is that neurotic and low self-esteem individuals feel more inadequate as partners (Peretti & Pedowski, 1997), and, as a consequence, feel more easily threatened by potential rivals.
Also, in the description of love styles suggested by Lee (1973) the mania style is characterized by a possessive attitude towards love and expressing jealousy towards the partner.The correlation between the mania love style and jealousy, measured by the Multidimensional Jealousy Scale, is of equivalent direction and similar intensity to the one reported in a study conducted by Attridge (2013).The mentioned study reported correlations between mania love style and cognitive, emotional and behavioral jealousy of .32,.37 and .32respectively.
All the correlations between romantic jealousy and self-esteem, neuroticism and the mania love style obtained in the current study correspond to previous findings on the relations between these constructs.
Additionally, when all personality variables are taken into consideration, it seems that personality profile of the people higher on jealousy, and any of its aspects, is opposite to the profile of healthy personality.

Conclusion
The verification of the psychometric characteristics of the Multidimensional Jealousy Scale on a Serbian sample confirms the findings of the original study and supports the validity of the Serbian version of MJS.The same goes for the internal consistency validity of the scale.
The results of the confirmatory factor analysis suggested that the original measurement model needs to be appended with a number of inter-item error covariance terms if an acceptable level of fit to the data is to be obtained.But, all the theoretical factor-item relations hold, generally supporting the internal validity of the scale on the Serbian population, albeit with the existence of interitem error covariances.The largest obtained deviation was with items 2 and 6 behaving as indicators of both cognitive and emotional jealousy factors.
Examination of factor structure of the instrument by gender showed that the tested models fit the female subsample, but not the male one.Closer examination revealed that this was primarily due to the difference in inter-item error covariances between the two samples, but that all theoretically expected factor-item relations remain practically the same in both subsamples.
External validity of MJS, examined by calculating correlations with a number of external variables, was also supported by the obtained results -PSIHOLOGIJA, 2017, Vol.50(4), 521-534 correlations with neuroticism were positive and statistically significant, with mania love style positive and moderate.Obtained correlations with self-esteem were negative and low, all in accordance with previous results, thus supporting the external validity of the scale on the current population.
It can be concluded that the obtained results support both internal and external validity of the Serbian version of MJS, and suggest that MJS can be used on the Serbian population as a valid measure of romantic jealousy constructs.

Figure 1 .
Figure 1.Model 1 -theoretical model of MJS latent structure with added error covariances.

Figure 2 .
Figure 2. Model 2 -modified model of MJS latent structure, two items added to the Emotional jealousy factor

Table 1
Internal consistency and descriptive statistics measures of scales used in the current study

Table 2
Fit indices for models presented in Figures1 and 2.
Fit indices: GFI -Goodness of Fit Index, AGFI -Adjusted Goodness of Fit Index, CFI -Comparative Fit Index, NFI -Normed Fit index, RMSEA -Root Mean Square Error of Approximation

Table 3
Item wise analyses of internal structure of MJS, descriptive statistics, alpha if item deleted and standardized regression weights for the two tested models of MJS latent structure.

Table 4
Correlation coefficients between the Multidimensional jealousy scale and assessed external variables