Emophilia: psychometric properties of the emotional promiscuity scale and its association with personality traits, unfaithfulness, and romantic relationships in a Scandinavian sample

The study addresses how frequently and easily individuals fall in love (emophilia) and evaluates whether emophilia is a robust and distinct construct linked to romantic outcomes. Drawing on evolutionary and psychological perspectives on romantic love and falling in love, the authors note limited, largely North American, research on emophilia and the need for replication and cross-cultural investigation. The purpose is to test the EPS’s psychometric properties in a Scandinavian sample, assess discriminant validity via associations with Big Five and Dark Triad traits, and test predictive validity for number of romantic relationships and unfaithfulness while adjusting for covariates. Hypotheses: H1: EPS is psychometrically sound with a two-factor structure and good internal reliability (alpha ≥ 0.80). H2: EPS shows satisfactory discriminant validity (r < 0.40 with personality traits). H3: Emophilia is positively associated with number of romantic relationships and unfaithfulness.

Romantic love is considered an adaptation facilitating pair-bonding, and falling in love varies with gender, culture, attachment, and personality. Emophilia integrates how easily and how often one falls in love and is related conceptually to sociosexuality, though oriented toward romantic feelings rather than sexual behavior. Prior research on emophilia is limited, predominantly led by Jones with North American samples, underscoring replication needs due to psychology’s replication crisis and potential cultural differences in perceiving romantic feelings. Prior associations with Big Five traits are mixed: Jones (2017) found no significant associations; Jones (2011) reported lower agreeableness, higher neuroticism among men, and higher extraversion among women linked to higher emophilia. Emophilia is positively associated with Dark Triad traits (Lechuga & Jones, 2021). Predictive associations reported include greater numbers of relationships, marriages/divorces, earlier engagements, multiple-partner pregnancies, infidelity, and uncommitted sexual relations (Jones, 2011, 2015; Pinto, 2015). These links motivate examining discriminant and predictive validity of the EPS in a Scandinavian context while accounting for covariates such as age, gender, and personality.

Design and sample: Cross-sectional online survey recruited via VG+ (Norway) and Aftonbladet+ (Sweden) articles about emophilia in October 2020. N = 2,607 (women 74.6%, men 24.7%, other 0.7%). Informed consent obtained. Measures: Romantic outcomes: single-item counts: lifetime number of romantic relationships and number of times unfaithful; response options 0–50 and 50+. Demographics: age (≤16, 17–89 in single-year steps, ≥90) and gender (man, woman, other). Dark Triad: Dirty Dozen (12 items; 1–9 Likert). Subscales (four items each) scored 4–36. Cronbach’s alpha: Machiavellianism 0.85, psychopathy 0.74, narcissism 0.81. Big Five: Mini-IPIP (20 items; 1–5). Subscales 4–20. Alphas: extraversion 0.80, neuroticism 0.72, openness 0.71, agreeableness 0.74, conscientiousness 0.65. Emophilia: Emotional Promiscuity Scale (EPS; 10 items, two factors: easily and often; 1–5 Likert). Composite 10–50, typically used due to high inter-factor correlation. Tabled descriptive stats reported; subscale alphas: easily 0.80; often 0.74; composite 0.85. Analysis: CFA with IBM Amos 27 to test two-factor EPS, allowing correlated errors between items 6 and 7 based on wording and prior findings. Factor loading cutoff > 0.32. Fit indices: CFI, TLI, RMSEA (90% CI), with acceptable thresholds of 0.90/0.90/≤0.10. Compared two-factor, hierarchical, bifactor, and single-factor models using AIC (lower is better) and fit indices. Tested measurement invariance with MGCFA across gender (men vs women) and age groups (≤35, 36–55, ≥56); configural fit via CFI/TLI/RMSEA; metric invariance via comparison of metric vs configural models. Correlations: Pearson’s r between emophilia and Big Five, Dark Triad, age; point-biserial for gender; p < 0.05 threshold; interpret 0.10/0.30/0.50 as small/medium/large. Predictive analyses: Overdispersion precluded Poisson; negative binomial regressions used with emophilia (z-scored) predicting counts of romantic relationships and unfaithfulness. Age included as an offset and also entered as a continuous adjustment variable; models estimated crude, adjusted for age/gender, adjusted for Big Five, adjusted for Dark Triad, and fully adjusted for all. Reported IRRs with 95% CIs; interpret IRRs of 1.22/1.86/3.00 as small/medium/large. Checked multicollinearity (VIF). Tested gender moderation via interaction terms.

Descriptives and reliability: EPS items 1–9 showed acceptable skewness/kurtosis (−2 to +2); item 10 deviated (skew < −2; kurtosis > +2). Internal consistency: easily α = 0.80; often α = 0.74; composite α = 0.85. CFA and model comparisons: Two-factor model fit: CFI = 0.95, TLI = 0.91, RMSEA = 0.072 (90% CI 0.067–0.078). Items 1–5 loaded on Easily; items 6–10 on Often; error terms for items 6 and 7 correlated r = 0.54; sensitivity analysis without this correlation marginally changed loadings (EP6 0.53→0.59; EP7 0.47→0.54). AICs: two-factor 556; hierarchical 592; bifactor 385; single-factor 1,435. Fit indices: hierarchical CFI 0.94, TLI 0.91, RMSEA 0.074 (0.069–0.080); single-factor CFI 0.85, TLI 0.76, RMSEA 0.120 (0.114–0.125); bifactor CFI 0.97, TLI 0.93, RMSEA 0.066 (0.060–0.073). Bifactor model was best fitting overall. Measurement invariance: MGCFA bifactor model across gender (CFI 0.97, TLI 0.94, RMSEA 0.042) and age groups (CFI 0.97, TLI 0.93, RMSEA 0.038) indicated moderate to good configural fit; metric invariance across age groups was not supported (p < 0.000). Correlations with traits and demographics (emophilia as focal): age r = −0.08**; extraversion r = 0.14**; agreeableness r = 0.06**; conscientiousness r = −0.17**; neuroticism r = 0.25**; openness r = 0.09**; narcissism r = 0.34**; psychopathy r = 0.14**; Machiavellianism r = 0.28**; gender r = 0.01 (ns). Effect sizes generally small; narcissism in the medium range. Predictive analyses (negative binomial): Number of romantic relationships: IRR (crude) 1.22 (1.18–1.26)***; adjusted for age/gender 1.20 (1.17–1.24)***; adjusted Big Five 1.21 (1.17–1.26)***; adjusted Dark Triad 1.19 (1.15–1.24)***; fully adjusted 1.21 (1.17–1.25)***. Unfaithfulness: IRR (crude) 1.43 (1.36–1.50)***; age/gender 1.40 (1.33–1.47)***; Big Five 1.37 (1.30–1.45)***; Dark Triad 1.30 (1.24–1.38)***; fully adjusted 1.25 (1.18–1.32)***. Effects were small; the fully adjusted effect for unfaithfulness was weaker than crude (non-overlapping CIs). VIF = 1.26; no multicollinearity. No significant gender moderation.

Findings support H1: the EPS demonstrated good internal reliability and a factor structure consistent with prior work; although two factors were evident, their high correlation supports use of a composite. Competing model tests favored a bifactor structure. H2 was supported: emophilia showed satisfactory discriminant validity against Big Five and Dark Triad traits, with correlations mostly small and below 0.40, except a medium association with narcissism, suggesting emophilia is not reducible to these traits. H3 was supported: emophilia predicted higher counts of romantic relationships and unfaithfulness with small effect sizes, consistent with the notion that ease/frequency of falling in love relates to romantic behaviors, though many other factors influence these outcomes. A discrepancy with prior work was noted for agreeableness (positive here vs negative previously), which may reflect cultural differences, sample or measurement differences. Cross-cultural comparability cannot be concluded given the absence of direct comparative tests. The cross-sectional, self-report design precludes causal inference; effects could be bidirectional or confounded, and statistical control can introduce bias if covariates are colliders or consequences of the focal variables. Overall, emophilia appears similar in Scandinavia to prior North American findings in structure and correlates, with some differences in effect magnitudes and specific traits.

The EPS exhibited good psychometric properties in a large Scandinavian sample, with evidence for a two-factor (and best-fitting bifactor) structure and satisfactory internal reliability. Emophilia demonstrated discriminant validity with personality traits and small positive associations with number of romantic relationships and unfaithfulness, indicating some predictive validity. Given the cross-sectional design and measurement limitations, definitive conclusions about causality and trait status are premature. Future research should employ longitudinal and cross-cultural designs, refine operational definitions of key outcomes, and examine additional determinants (e.g., attachment, cognitive factors, life events) to clarify whether emophilia is a stable, unique personality trait or a behavioral outcome of multiple interacting influences.

Key limitations include: (1) Cross-sectional, self-report design susceptible to recall bias, social desirability, and common method bias; causality and directionality cannot be inferred. (2) Convenience sampling via online, subscriber-only newspaper platforms may induce selection bias and limit generalizability (e.g., underrepresentation of older adults or those with fewer financial means, and individuals especially interested in emophilia). (3) Lack of standardized definitions for unfaithfulness and romantic relationship likely introduced idiosyncratic interpretation, potentially inflating associations. (4) Measurement invariance across age groups was not supported at the metric level, limiting comparisons across ages. (5) Potential omitted variables (e.g., attachment style, cognitive factors, life events) that may explain emophilia and outcomes. (6) Use of statistical controls may obscure or bias estimates if covariates are consequences or colliders.