Extending the prediction of infidelity using a five-factor model



Journal Title

Journal ISSN

Volume Title



This research examined the contributions of individual, general relationship, and sexual variables in predicting different types of infidelity. This task was accomplished by examining correlations, as well as the contribution of clusters of variables to initial modeling of each type of infidelity and the relative predictive strength of certain variables to different types of infidelity. The overall goal of this research was to provide evidence that behaviors not traditionally thought of as emotional or sexual infidelity, but which are nevertheless considered unfaithful, are predicted differentially from more traditional forms of emotional and sexual infidelity. It may be that these different predictive constructs are in line with Allen et al.’s (2005) conception of different predictive variables at different points in the development of relationship infidelity. It may also be the case that there are different timelines, an understanding of which could be aided by the current research.

A pilot study and initial study (Study 1) involved item generation and exploratory factor analysis to develop the initial Infidelity Scale. It was hypothesized that infidelity would be a multidimensional construct and that construct validity would be established by showing that the infidelity subscales correlated with other relationship constructs in expected ways. The hypotheses were supported, generating a five-factor Infidelity Scale consisting of 37 items. Study 2 confirmed the five-factor structure of the Infidelity Scale and provided further evidence for construct validity. Confirmatory factor analysis largely supported the hypotheses. However, due to lack of fit with the hypothesized model, two items were deleted from the scale. Confirmatory analysis was run again and yielded an acceptable model fit to the data.

The current study employed data from 487 college students. The final factor structure of the Infidelity Scale was confirmed. Simultaneous Linear Regression analyses were used to examine the relative contributions of several clusters of variables to predicting the Infidelity subscales. Hierarchical Multiple Regression was used to look at the interactions of these sets of variables with gender. Steps taken toward modeling the subscales revealed that the individual sexual variables emerged as the primary predictor for all of the Infidelity subscales. Also, all subscales that involved some degree of sexual act produced the same sequence of predictors. Gender interacted with the individual sexual variables in predicting Sexual Infidelity and Deviant Involvement such that these variables were more predictive of these subscales for males. Overall, these findings provide evidence that different models for the infidelity subscales may exist. However, subscales with more emotional/cognitive content may have similar models, and subscales with more physical content may have similar models. Additionally, it is possible that there are different models for predicting the subscales by gender. These findings are also discussed in terms of Allen et al.’s proposed temporal sequence. Limitations of this inquiry and directions for future research are discussed.



Infidelity, Romantic relationships