Construction and Validation of the Attitude Toward Returning to an Ex-Partner Scale

Abstract

Background/Objectives: When a relationship ends due to abuse, a favorable attitude toward reconciliation may become a risk factor. The objective of this study was to develop and validate an instrument to measure the attitude toward returning to an ex-partner. Methods: A pilot study was conducted to evaluate the dimensionality and psychometric quality of the items. The main study involved 55 women who had been victims of gender violence. Results: Following item analysis and assessments of reliability (α = 0.93) and validity, a unidimensional 16-item scale was developed. The instrument, named the “Attitude Toward Returning to an Ex-partner Scale” (ATRES), allows for the identification of predispositions to return to a relationship in which serious abuse has occurred. Moreover, the findings revealed that a heightened perception of danger, along with forgiveness directed toward oneself, the other person, and the situation, was associated with a less favorable attitude toward reconciliation. Conversely, high religiosity predisposed individuals to rekindle the relationship. Conclusions: The scale could serve to facilitate interventions, mainly in situations where restoring the relationship can be a risk. The assessment of the predisposition to forgive the ex-partner—namely, the individual who perpetrated the abuse—as well as the victim’s attitude toward re-engaging in the relationship, constitute important considerations for preventing revictimization. The ATRES is the first self-report measure designed to assist researchers and professionals in the precise assessment of specific beliefs and myths underlying the reinstatement of a relationship.

1. Introduction

There are several reasons why relationships come to an end. From both clinical and psychoeducational perspectives, it is important to distinguish between relationships in which reconciliation is possible—when grievances can be addressed and repaired—and those in which the harm is irreparable from the victim’s point of view. In such cases, encouraging reconciliation may not be advisable, as it could undermine the victim’s recovery and emotional safety. This may occur in the context of relationships involving gender-based violence (Labrador 2015; Prieto-Ursúa 2015; Sengupta et al. 2024). From this perspective, the goals of therapeutic interventions should be different. There are interventions that focus on resolving conflicts, promoting the capacity to forgive the transgressor within the couple, and restoring the relationship. Other interventions are aimed at helping victims of gender-based violence overcome the guilt—by facilitating self-forgiveness—that arises as a consequence of the abuse (Naismith et al. 2022). To achieve this, it is essential to strengthen their self-esteem and self-confidence, enhancing autonomy, address the negative feelings linked with the perpetrator, and help them to overcome the psychological consequences of the harm suffered without these actions necessarily leading to reconciliation with the partner (Sengupta et al. 2024).

Among the types of irreparable damage that can occur in relationships, gender violence is prominent. Exposure to this type of violence is associated with a series of negative consequences for women (Piccinini et al. 2023). In most cases, the cognitive and emotional effects of the abuse persist even after the relationship is over (e.g., low self-esteem, self-blame, learned helplessness, anxiety, depression). These consequences may increase victims’ vulnerability and potentially influence their attitude toward resuming the relationship with the perpetrator (Ford-Gilboe et al. 2023). In addition to these psychological factors, lack of social support, financial hardship, and unemployment are critical contributors to the likelihood of returning to an abusive relationship (Delegación de Gobierno para la Violencia de Género 2019). With consideration of the phases described by the “cycle of violence” theory (Walker 1979), reconciliation could take place within the so-called “honeymoon phase” in which promises and expectations of a change can generate a favorable attitude toward resuming the relationship. Moreover, the perpetrator’s apologies play a crucial role in fostering forgiveness in the victim, as they may enhance perceptions of the relationship’s value. However, this influence on forgiveness tends to weaken as the severity of the abuse increases (Forster et al. 2021).

Other research has confirmed the negative effect that alternating positive and negative behaviors by the aggressor has on the victim’s self-esteem, attachment bonds, and interdependence (Credidío et al. 2024; Fuertes 2023). Therefore, in cases in which the violence has been relatively moderate and intermittent, such that the perception of danger is low and the abuse may be considered justified, the victim could believe in the promises of the aggressor to change and thus try to reestablish the relationship. Moreover, internalization of the cognitive schemes developed during the relationship as a way to endure and make sense of the situation (justification of the partner’s behavior, attribution of blame, minimization of abuse, etc.) can persist over time and increase vulnerability (Delegación de Gobierno para la Violencia de Género 2019; Badenes-Sastre et al. 2024).

Attitudinal variables have also been associated with difficulty in breaking off an abusive relationship and to feelings of failure when it is not possible to reestablish the bond. Specifically, people with traditional attitudes regarding the role of women, high religiosity, and the moral commitment associated with marriage tend to maintain relationships even under risky circumstances (Amor and Echeburúa 2010; Green et al. 2023; Heron et al. 2022).

In addition to the abovementioned factors, forgiveness is relevant for understanding the reactions to abuse in interpersonal relationships (Sengupta et al. 2024; Vázquez et al. 2017). People exposed to violent situations of any type frequently experience emotions such as fear, guilt, or sadness. They may also witness their socio-familial and affective relationships deteriorating and can suffer various physical ailments (Lutwak 2018; Nathanson et al. 2012). Forgiveness has been used as a strategy to reduce unpleasant emotions, depression, anxiety, stress, anger, neuroticism, rumination, fear, and hostility (Cao et al. 2025). Additionally, it has been found to contribute to a reduction in the incidence of risk behaviors (such as alcohol, tobacco, or other substance use) (Tala and Valenzuela 2020). In many contexts, forgiveness is the key to reconciliation, although this depends on the seriousness of the offense and on whether there is certainty that the aggression will not be repeated (Forster et al. 2021). However, forgiveness and reconciliation are not identical, and the presence of one does not necessarily imply the presence of the other. In situations in which restoration of the interpersonal relationship is not the goal, forgiveness pertains not only to overcoming hate and resentment toward the aggressor (forgiveness of others) but also to reconciliation with oneself (forgiveness of oneself or intrapersonal forgiveness) and the situation, which can release the victim from self-blame (Cao et al. 2025; Enright 1996; Song et al. 2025). According to Worthington (Worthington 2023, 2024), forgiveness does not equate to justification of the other’s behavior, does not necessarily imply reconciliation, and does not represent consent to the aggression.

Ultimately, when it comes to forgiveness, its benefits for mental health and well-being, as well as its positive effects on socio-emotional development, have been well-documented in numerous studies (Cao et al. 2025; Gao et al. 2022; Rodríguez et al. 2018). Specifically, in the context of gender-based violence, self-forgiveness involves minimizing the feelings of guilt that are often associated with a woman’s self-perceptions, past decisions, emotional ties to the aggressor, and internalization of messages received from the perpetrator, family members, and society (Cheng and Yim 2008). Moreover, situational forgiveness involves gaining a deeper understanding of what occurred and disengaging from the negative emotions that undermine the victim’s well-being and ability to move forward (Thompson et al. 2005). Despite this, some authors have concluded that in certain situations, promoting the ability to forgive—particularly the aggressor—from a traditional and limited view can have detrimental effects on the victim, increasing the favorability of the attitude toward resuming the relationship with the aggressor and raising the probability of being victimized again (McNulty 2010; McNulty 2011). This can worsen the self-concept (Luchies et al. 2010) and perpetuate dynamics of abuse that would put the individual at risk again (Amor and Echeburúa 2010; Gordon et al. 2004).

Considering the above, the consequences of the relationship being broken must also be analyzed from a professional point of view to facilitate interventions. Restoring psychological, affective, and social functioning, among other types, must be the priority, as well as mitigating future possible risks for the victim (Lopes-Dos-Santos and Anaut-Bravo 2024). To achieve this, assessing the predisposition to forgive the ex-partner, i.e., the individual who perpetrated the abuse, and the victim’s attitude to returning to the relationship, are relevant considerations for interventions. However, there are no standardized self-report measures that allow for the precise assessment of this construct or assist professionals in drawing comparative conclusions on the matter. A review of the literature identified only two instruments that measure cognitive, behavioral, and affective attitudes toward ex-partners: the Ex-Partner Attitude Scale by Banse and Kowalick (Banse and Kowalick 2007) and its later adaptation by Imhoff and Banse (Imhoff and Banse 2011). These studies also incorporate implicit measures of attitudes toward ex-partners based on priming tasks. However, the measures provided by these instruments and tasks do not specifically assess the attitude toward rekindling the relationship. Considering this, the objective of the present study was to develop and validate a scale that measures the predisposition to return to the ex-partner given the absence of instruments that specifically evaluate this construct.

2. Pilot Study

2.1. Materials and Methods

2.1.1. Participants

The sample of the pilot study consisted of 212 college students enrolled in a psychology degree program at the University of Jaén (Spain). Of the participants, 76.9% were female and 23.1% were male. The age ranged from 18 to 50 years (M = 22.01 years, SD = 5.17). Regarding marital status, most of the participants were single (93.9%); few were separated or divorced (6.1%). Regarding religion, 68.9% of the participants declared themselves to be Catholic, 21.6% agnostic, 8.7% atheist, and 0.8% Jehovah’s Witness. Regarding religious practice, the majority of the participants considered themselves non-practicing (78.8%; practicing = 21.2%).

2.1.2. Measures

We used a sociodemographic questionnaire. A questionnaire was designed to obtain information about the following variables: age, gender (female, male, other), university degree and course, marital status (single, separated, divorced, other), type of religion (open question), and degree of religiosity (1 = low, 2 = moderate, 3 = high).

We also used the Attitude Toward Returning to an Ex-partner Scale (ATRES; initial version, 25 items).

2.1.3. Procedure

Development of Scale Items

To develop items for the ATRES, the principles of representativeness (extent to which the item reflects the theoretical domain of the construct), relevance (degree to which the item is essential and meaningful to measure the construct), comprehensibility (how clear, unambiguous, and understandable the item is to the target population), and simplicity (whether the item is formulated in a concise and straightforward manner) were followed (Boateng et al. 2018; Rubio et al. 2003).

A total of 56 initial items were developed based on empirical findings and theoretical discussions related to attitudes toward returning to abusive relationships. These attitudes—which include cognitive, emotional, and motivational components—have been documented in the literature as playing a significant role in victims’ decisions to either break away from or return to violent partners. Prior research has identified several key beliefs associated with this tendency, including feelings of regret, the hope for change in the abuser, the belief that the abuser is a good parent or family member, internalized romantic myths, and social pressure to maintain the relationship (Naismith et al. 2022; Walker 1979). Conversely, other attitudes reflect resistance to reconciliation, often linked to perceptions of danger, awareness of the damage suffered, and emotional distancing from the aggressor.

The items in the ATRES were designed to capture this spectrum of attitudes. Some items were scored such that higher values indicated a more favorable attitude toward returning to the ex-partner, while others were reverse-scored, reflecting an unfavorable stance. The former addressed justifications or cognitive rationales for reconciliation, often connected to the cycle of violence and emotional ambivalence. The latter reflected protective mechanisms and awareness of harm. A 5-point Likert-type response format was used, ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). To ensure the validity of the items, an assessment was made by three professional experts in this field. Specifically, they were asked about the following aspects: (a) the representativeness or adequacy of the items with respect to the subject of this study, (b) the clarity and comprehensibility of the wording, and (c) the sufficiency and diversity of the items with respect to the subject of this study. Items not meeting these criteria were eliminated; 25 items remained after this review process and were tested through a pilot study.

Procedure for the Study

To be involved in the analysis of the psychometric properties of the preliminary version of the tool, participants had to meet the following two criteria: (a) being without a partner at the time of data collection, and (b) having previously had a relationship at least 3 months in duration. After receiving the instructions and reading and signing the consent form, the participants responded to the sociodemographic questionnaire and the 25 initial ATRES items. In the final section, questions regarding the degree of understanding of the items and any errors detected were included, along with space for suggestions. None of the items were reported as being inappropriate or incomprehensible.

Regarding ethical considerations, this research was previously approved by the Ethics Committee for research involving humans of the University of Jaén (Spain). No minors participated in this study. The data were anonymized in accordance with Organic Law 3/2018 (5 December) on the Protection of Personal Data and guarantee of digital rights.

2.2. Analysis Plan

For the analysis of the psychometric quality of the items, the discrimination index and the differential functioning of each item were considered. Specifically, the assumption of a normally distributed sample was verified through analysis of the distribution of the items (asymmetry and kurtosis), with calculation of mean (M) and standard deviation (SD) values. Items with asymmetry and kurtosis values of ±1, weighted means between 1.5 and 4.5 points, and an SD ≥ 1 (Ferrando and Anguiano-Carrasco 2010) were considered adequate. Collinearity was assessed through bivariate correlations between the items and their differential functioning.

The dimensional structure of the scale was also explored via exploratory factor analysis (EFA) and, later, confirmatory factor analysis (CFA) with a separate sample. For these analyses, the FACTOR program (version 10.8.04) was used. For the EFA, items with factorial loadings >0.400 were considered. The CFA was applied to verify the unifactorial model resulting from the EFA using a new and separate sample and the EQS statistical package. The maximum likelihood estimation method was used, along with the corrections of Satorra and Bentler (Satorra and Bentler 2001). To evaluate the fit of the model, the Chi-square statistic (χ²) and root mean square error of approximation (RMSEA; values ≤ 0.08 indicate a good fit), comparative fit index (CFI), and adjusted goodness-of-fit index (AGFI) values were calculated. For a good model fit, the CFI values must be ≥0.95 (Byrne 2013). The internal consistency was calculated using the Omega coefficient (FACTOR program, v. 10.8.04).

Finally, to assess construct validity, the relationships of the scale scores with sociodemographic variables (sex, age, and degree of religiosity) were analyzed. Differences in ATRES scores between categorical variables were analyzed using ANOVA. Post hoc comparisons of variables showing homogeneity of variance were performed using the Tukey test due to its versatility and ability to control for type I error (Abdi and Williams 2010). In cases with inequality of variance and similar potentialities, the homonymous Games–Howell test was used (George and Mallery 2001).

2.3. Results

2.3.1. Analysis of Scale Items

Table 1. Average scores and standard deviations of the final items of the Attitude Toward Returning to an Ex-partner Scale.

	M	SD	Asymmetry	Kurtosis	Correlations
Item 1	2.41	1.27	0.52	−0.67	0.760 **
Item 2	3.17	1.26	−0.21	−0.85	−0.144 *
Item 3	3.00	1.34	−0.20	−1.00	0.523 **
Item 4	3.00	1.34	−0.14	−1.17	0.705 **
Item 5	2.40	1.27	0.43	−0.91	0.766 **
Item 6	2.58	1.36	0.26	−1.12	0.759 **
Item 7	2.85	1.33	0.03	−1.06	0.547 **
Item 8	3.49	1.25	−0.58	−0.45	−0.007
Item 9	2.41	1.28	0.48	−0.84	0.813 **
Item 10	3.89	1.18	−0.98	0.19	0.078
Item 11	2.51	1.32	0.33	−1.03	0.632 **
Item 12	3.06	1.34	−0.01	−1.05	−0.535 **
Item 13	2.04	1.07	0.72	−0.20	0.727 **
Item 14	3.10	1.28	−0.04	−0.93	−0.206 **
Item 15	2.28	1.25	0.55	−0.80	0.827 **
Item 16	2.17	1.29	0.75	−0.58	0.791 **
Item 17	2.02	1.11	0.89	−0.07	0.759 **
Item 18	3.26	1.36	−0.19	−1.04	−0.469 **
Item 19	2.57	1.34	0.09	−0.94	0.707 **
Item 20	2.34	1.15	0.40	−0.62	0.613 **
Item 21	1.65	1.00	1.42	1.35	0.586 **
Item 22	2.04	1.09	0.86	0.12	0.800 **
Item 23	3.18	1.30	−0.23	−0.89	−0.063
Item 24	3.75	1.22	−0.75	−0.31	0.169 *
Item 25	3.55	1.27	−0.40	−0.89	−0.416 **

** The correlation is significant at the 0.01 level (bilateral). * The correlation is significant at the 0.05 level (bilateral).

provides descriptive statistics for the items. The weighted mean of the items was between 1.5 and 3.75, and all of them had SD values ±1. The distribution analysis showed that all the items had excellent (±1) or acceptable (±1.5) asymmetry and kurtosis. The absence of multicollinearity was verified by inspecting the correlations between items, all of which were <0.90 (George and Mallery 2001). Items 8, 10, and 23 were omitted because they did not show a significant relationship.

2.3.2. Dimensionality and Reliability

Parallel analysis showed that the optimal number of factors, or solutions, was one. The initial solution of the EFA (maximum likelihood method with diagonal rotation), performed without factor fixation, explained 36.4% of the total variance in the scores. In a subsequent analysis, a model with six factors explaining 64.4% of the variance was derived. Despite its high explanatory power, however, this solution was not used given the unequal saturation of the items. Specifically, of the total of 25 items, 16 were saturated in the first factor, 4 in the second, and only 1 in the third and fourth. None of the items were saturated exclusively in factors 5 and 6. The free factor analyses were repeated until the smallest final model was obtained, which explained 62% of the total variance and in which all items had a saturation value > 0.400 and were distributed among four factors (16 items in the first, 3 in the second, and only 1 in the third and fourth factors). Given that most of the items were saturated in the first factor and that there was no difference in terms of the theoretical content of the items saturated in the remaining factors, we forced a one-factor solution. This yielded a factorial solution that explained 37.1% of the total variance, with seven items (2, 8, 10, 14, 18, 23–24) having a load < 0.400. Finally, this analysis was repeated after omitting these items, resulting in a solution that explained 51% of the total variance via a single factor, in which all items (1, 3, 4, 5–7, 9, 11, 13, 15–17, 19–22) had saturation values > 0.503. The items and single-item solution can be seen in

Table 2. Saturation of the items in the final one-factor solution.

Items	F 1
(1) 1. When he/she asks to speak to me, I cannot refuse to do so.	0.763
(3) 2. No matter how many presents he/she gives me, I would not return to the relationship.	0.503
(4) 3. I could give my ex-partner another chance if he/she treated my family well.	0.697
(5) 4. Whatever he/she said, I wouldn’t go back to my ex-partner.	0.766
(6) 5. If he/she respected my decisions or choices, I could give him/her another chance.	0.756
(7) 6. If he/she was good to me, I would go back to him/her.	0.526
(9) 7. I do not wish to return to my ex-partner under any circumstances.	0.817
(11) 8. It is not in my plans to resume this relationship, no matter how much he/she says that he/she accepts me as I am.	0.622
(13) 9. If he/she asked me for another chance, it would be difficult for me to refuse.	0.739
(15) 10. I would go back to him/her if he/she expressed remorse about what he/she has done.	0.836
(16) 11. I don’t want to give him/her any more chances.	0.799
(17) 12. The fact that my ex-partner is a good mother/father does not imply that we should get back together.	0.770
(19) 13. Even if he/she regretted his/her actions, I would not return.	0.705
(20) 14. If he/she sent me flowers or gifts, I could go back to him/her.	0.607
(21) 15. I would not consider going back to my ex-partner, even if he/she behaves well towards me.	0.596
(22) 16. If he/she told me that he/she cannot live without me, I would go back to him/her.	0.806

. The items are provided in Spanish and English in Appendix A and Appendix B, respectively.

To guarantee an optimized solution, a CFA was carried out with a separate sample of 202 participants. The results were as follows: χ² = 156.32, df = 34, CFI = 0.91, RMSEA = 0.05 (0.04, 0.06), χ² = 146.331, robust CFI = 0.92, and robust RMSEA = 0.04 (0.03, 0.05).

Internal consistency analysis yielded an excellent value (ω = 0.92). No improvement was found with the elimination of any item.

2.3.3. Construct Validity

The scale scores according to the sociodemographic variables of age, sex, and degree of religiosity were analyzed. The correlation between age and ATRES score was negative, r = −0.146, p < 0.05, i.e., as age increases, the intention to return to the ex-partner decreases. Regarding sex, no mean differences were found in ATRES scores between women and men, F(1, 211) = 1.85, p = 0.171. Finally, a one-way ANOVA examining ATRES scores across levels of religiosity yielded a significant effect, F(8, 211) = 3.08, p = 0.002. Post hoc analysis revealed that the high-religiosity group showed a more favorable attitude toward returning to the ex-partner than the low-religiosity group, M_low = 2.21, SD = 0.75, M_high = 3.00, SD = 1.25.

3. Main Study

3.1. Materials and Methods

3.1.1. Participants

The main study included 55 women who had been victims of gender violence. The age range was from 18 to 68 years (M = 40.4 years, SD = 10.5). Regarding sociodemographic characteristics, 10.9% of the participants had not completed primary studies, 32.8% had completed primary studies, 43.6% had attended high school, and 12.7% were attending or had completed higher education. Regarding religion, 78.2% of the participants reported being Catholic, 3.6% were Jehovah’s Witnesses, 1.8% were Muslims, and 16.4% did not answer this question. Regarding religious practice and the degree of religiosity, 40% of the participants considered themselves practicing, of whom 60% stated that they had a medium or high degree of religiosity.

3.1.2. Measures

Using the sociodemographic questionnaire, the following data were collected: age, educational level, marital status (single, separated, divorced, other), religiosity (open question), actively practicing a religion (yes/no), and degree of religiosity (1 = low, 2 = moderate, 3 = high).

Attitude Toward Returning to an Ex-partner Scale (ATRES): the final version of the ATRES is a 16-item scale with 5-point Likert-type response options (from “totally disagree” to “totally agree”). The reliability coefficient was α = 0.950.

The Perception of Danger and Second Victimization Scale (Aranda López et al. 2014) consists of 20 items with Likert-type response options ranging from 1 (totally disagree) to 7 (totally agree). This instrument has four dimensions: (1) perception of danger, (2) police dissatisfaction, (3) judicial dissatisfaction, and (4) health dissatisfaction and social networks. For the present study, the first dimension was used because it is relevant to the research objectives. The reliability coefficient of the 6-item subscale was α = 0.900.

The Heartland Forgiveness Scale (HFS) (Thompson et al. 2005) was translated into Spanish and validated by Montes-Berges and Vázquez (Montes-Berges and Vázquez 2014). This version of the instrument, used herein, comprises 18 items and assesses a single dimension: the willingness to forgive oneself, others, and situations beyond individual control. After inverting the reverse-scored items, higher scores reflect more forgiveness. The response options are in a 7-point Likert-type format, where 1 denotes “almost always false about me” and 7 “almost always true about me”. The reliability coefficient was α = 0.790.

3.1.3. Procedure

The main study involved women who had experienced gender violence (in a couple relationship at least 3 months in duration) and who were not in a couple relationship during the study period. Participants answered the questionnaires on their own during one of the initial sessions of a therapeutic program provided by the Andalusian Institute for Women. Initially, the objective of this study and other relevant aspects (anonymity and voluntariness) were presented via the informed consent form. The instruments were presented in a counterbalanced order using the Latin square method (Shadish et al. 2002). Prior to conducting this study, the protocol was approved by the Ethics Committee of the University of Jaén; in addition, permission was obtained from the center and all participants. The individual in charge of explaining and administering the questionnaires was the main therapist of the program.

3.2. Analysis Plan

The psychometric properties of the instrument were evaluated through analysis of the individual items (see the “analysis plan” section of the pilot study for details). The convergent validity of the ATRES, Perception of Danger and Second Victimization Scale, and HFS and the construct validity of the sociodemographic variables were calculated.

3.3. Results

3.3.1. Item Properties

Regarding the properties of the items,

Table 3. Mean scores and standard deviations of the final items of the EAVE in the main study.

	M	SD	Asymmetry	Kurtosis	Correlations
Item 1	1.87	1.5	1.3	0.37	0.632 **
Item 2	2.16	1.64	0.137	−0.84	0.496 **
Item 3	1.65	1.84	1.432	1.46	0.771 **
Item 4	1.71	1.73	1.315	1.22	0.891 **
Item 5	1.62	1.66	1.477	1.07	0.887 **
Item 6	1.51	1.36	1.190	1.47	0.844 **
Item 7	1.62	1.46	1.451	1.31	0.891 **
Item 8	1.62	1.41	1.461	1.47	0.766 **
Item 9	1.50	1.50	1.232	1.30	0.887 **
Item 10	1.56	1.45	1.419	1.21	0.895 **
Item 11	1.84	1.87	1.268	0.29	0.818 **
Item 12	1.60	1.73	1.868	0.97	0.859 **
Item 13	1.95	1.59	1.455	1.39	0.823 **
Item 14	1.95	1.67	1.368	1.3	0.723 **
Item 15	2.17	1.61	1.315	−0.84	0.586 **
Item 16	1.96	1.52	1.451	1.46	0.727 **

** The correlation is significant at the 0.01 level (bilateral).

shows the descriptive statistics and discrimination index for all items. All items showed an adequate relation with the total score.

3.3.2. Convergent Validity

The convergent validity was determined through analysis of the correlations with the “perception of danger” subscale and HFS. The results showed a significant negative relationship between the ATRES and HFS scores, r = −0.245, p = 0.043. Participants showing greater willingness to forgive the situation, the other person, and themselves also expressed less willingness to return to the abusive ex-partner. The relationship with perception of danger showed the same pattern, r = −0.156, p = 0.038, i.e., a greater perceived risk on the part of the victim was associated with less willingness to resume the relationship and to forgive (

Table 4. Relationships between the attitude toward returning to the ex-partner, perception of danger, and attitude toward forgiveness.

	1	2	3
ATRES	1
Perception of danger	−0.156 *	1
HFS	−0.245 *	−0.302 **	1

** The correlation is significant at the 0.01 level (bilateral). * The correlation is significant at the 0.05 level (bilateral).

).

3.3.3. Construct Validity

Regarding the sociodemographic variables, there were no significant correlations of the attitude to return to the ex-partner with age, r = −0.195, p = 0.153, or the type of religion practiced, r = −0.098, p = 0.477, but there was a significant correlation with the degree of religiosity, r = 0.262, p = 0.024.

4. Discussion

Interventions for people who have been harmed by their partners or ex-partners may consider the consequences of relationship restoration when this could entail some risks (Montes-Berges and Vázquez 2014; Sengupta et al. 2024). However, there is no standardized tool to accurately assess the victim’s willingness to resume the relationship or the underlying reasons. The objective of the present study was to construct and validate the newly developed ATRES. This instrument was specifically designed for evaluating aspects of gender violence, although it could be adapted to other facets of couple breakups (e.g., infidelities, lack of commitment, addictions, or other conflicts).

As a result of the item analysis and the examinations of construct validity, reliability, and factorial structure, the final instrument comprised 16 items loading onto a single factor, with responses recorded on a five-point Likert-type scale. The indices used for psychometric adjustment indicate that it is a valid and reliable instrument. The ATRES allows the attitude toward resuming the relationship to be assessed, where higher scores reflect greater willingness to return to the ex-partner. The items pertain to the reasons why (according to the literature and clinical practice) victims of domestic violence may return to an abusive ex-partner. These arguments are formed and consolidated as a result of experiences endured during the relationship (Delegación de Gobierno para la Violencia de Género 2019; Badenes-Sastre et al. 2024; Amor and Echeburúa 2010). These beliefs are also influenced by the normalization of gender-based violence, which is rooted in cultural beliefs and values that, to some extent, sustain and justify its perpetration (Garrido-Macias et al. 2017). This underscores the importance, within the context of interventions for victims, of considering cognitive processes. Specifically, it is essential to work collaboratively with the individual to modify negative cognitions that contribute to the decision to enter into or remain in an abusive relationship. Furthermore, it is crucial to promote the recognition of one’s self-worth, emphasizing that being in a relationship does not define one’s inherent value.

The convergent validity (external validity) analysis revealed an inverse relationship between the attitude toward returning to the ex-partner and perception of danger. Although research on the processes that follow the termination of a conflictive relationship is very scarce, findings regarding interpersonal and intrapersonal processes occurring during the relationship allow us to interpret the results. The severity of the abuse influences individuals’ responses to conflict, including whether or not they ultimately decide to continue with the relationship (Montes-Berges and Vázquez 2014; Garrido-Macias et al. 2017). According to the results obtained in the present study, this is also applicable after the person has left the relationship. In general, empirical evidence has shown that an individual will be more likely to leave a relationship when the perceived severity of the transgression is high (Montes-Berges and Vázquez 2014); likewise, they will be less likely to resume the relationship if they believe that the risk will persist (Sengupta et al. 2024).

As discussed in the Introduction section, one of the most complex aspects of gender violence is that even when it is frequent and severe, some women choose to remain with their partner for many years and return to them even if they have been able to leave them temporarily. Factors such as fear, lack of social support and resources, and economic instability are key to understanding this situation of vulnerability (Delegación de Gobierno para la Violencia de Género 2019). Moreover, other aspects related to the psychological consequences of abuse may also be involved. In this context, leaving an abusive relationship is a process that may encompass periods of denial, guilt, and suffering before women are able to recognize the reality of the abuse; this usually marks the beginning of separation and recovery from the abuse suffered (Rodelli et al. 2022; Metts and Cupach 2007). Importantly, beyond these individual and relational dynamics, research highlights the role of structural features that shape the context in which IPV occurs. These include broader gender inequalities, the availability and enforcement of legal protections, women’s economic insecurity, the persistence of social norms that normalize violence, and institutional weaknesses in health, justice, and protection systems (Delegación de Gobierno para la Violencia de Género 2019; Amor and Echeburúa 2010; Badenes-Sastre et al. 2024; Heron et al. 2022). Addressing IPV therefore requires not only psychosocial and individual-level interventions but also structural changes aimed at reducing gender inequities, strengthening legal frameworks, and transforming social norms that perpetuate violence.

In this sense, forgiveness could be key to understanding and predicting attitudes that predispose an individual to return to their ex-partner (Vázquez et al. 2017; Heise and Garcia-Moreno 2002; Vaca Ferrer et al. 2020). The results of this research showed a negative relationship between these variables: the greater the capacity to forgive, the less positive the attitude toward resuming the relationship. Understanding this finding requires adoption of a concept of forgiveness which takes account of the need to overcome damage and, fundamentally, guilt (Cao et al. 2025; Song et al. 2025; Crapolicchio et al. 2021; Záhorcová et al. 2023).

Regarding partner violence, studies have revealed that experiences of abuse and violence may affect the victim’s cognition (Ford-Gilboe et al. 2023; Badenes-Sastre et al. 2024). The need to adapt to the aversive situation is characterized by, among other things, the appearance of cognitive distortions such as minimization, denial, or dissociation. In addition, the way in which the victim perceives themselves, others, and the world changes. Such changes may entail an impoverished self-image and the appearance of feelings of guilt and shame for having been mistreated (Beristáin 2004; Brémault-Phillips et al. 2022). In this respect, developing the ability to forgive oneself, the situation, and even the perpetrator seems to be a protective factor since it reduces the willingness to restart the relationship. Forgiveness appears to constitute a viable strategy for coping with experienced harm (Echeburúa et al. 2001). When an individual acknowledges the harm suffered, a cognitive and emotional shift occurs—moving from hatred and resentment to more positive or neutral emotions—which plays a fundamental role in the assimilation of the event and the alleviation of emotional pain (Rocha et al. 2017).

Focusing specifically on the willingness to engage in self-forgiveness, this type of forgiveness is understood to be the result of a deliberate and voluntary process involving emotional and attitudinal change, ultimately leading to the release of self-blame (Bernard and Cassady 1957; Cao et al. 2025). Various studies on self-forgiveness suggest that reducing resentment and replacing negative thoughts with attitudes of self-acceptance not only protects against the impact of past harm but also promotes self-love, tolerance of personal mistakes, and self-compassion (Worthington 2023). This may be explained by the observed relationship between self-forgiveness and increased levels of self-esteem, personal confidence, and psychological well-being, as well as improved conflict resolution abilities (Song et al. 2025; Fariña et al. 2020).

Finally, analysis of the ATRES scores in victims of gender violence according to sociodemographic variables, religiosity, and age showed that only the degree of religiosity affected the attitude toward reuniting with the ex-partner. Specifically, participants with higher scores on this item showed more favorable attitudes toward resuming the relationship. According to the literature, religion generally plays a vital role in shaping individuals’ beliefs about themselves, the world, and their understanding of both everyday and extraordinary events (Mróz et al. 2024). People often rely on their meaning-making systems for guidance and purpose in stressful situations (Park and Hale 2014). In light of this, it is plausible that in the case of women who are victims of gender-based violence, their religious belief systems may mediate the interpretation of their experiences and guide their decisions. Specifically, some authors suggest that religiosity in victims is one of the most influential factors in the decision to remain in or leave an abusive relationship (Ríos 2009). This role of religiosity appears to be linked to the emphasis on forgiveness—sometimes from a perspective of passivity and submission—in response to transgressions. Moreover, it tends to reinforce a moral commitment to marriage. Both of these aspects may ultimately constrain a woman’s decision to leave or return to a violent relationship (Amor and Echeburúa 2010; Green et al. 2023; Heron et al. 2022).

This study had some limitations. The differences in characteristics between the pilot and main study participants makes it difficult to generalize the results. Thus, the psychometric properties of the items could be questioned, including the dimensionality, validity, and reliability. The results obtained with the university students may not be applicable to female victims of gender violence. However, the reliability and validity results were largely replicated in the main sample. Another limitation concerned the number of participants recruited to the main study. Although about five participants per item is recommended (Carretero-Dios and Pérez 2005), i.e., a total of 80 people in our case, the vulnerable position of female victims of gender violence means that access to them is restricted and can be complicated.

In future research, the attitude construct could be studied together with other relevant constructs cited in the literature with respect to abuse in the couple relationship, such as subjective commitment, included in the Investment Model proposed by Rusbult and Buunk (Rusbult and Buunk 1993) and replicated in more recent studies (Gilbert and Gordon 2017; Mikkelson and Ray 2024). Likewise, considering the influence of the type of abuse experienced, the duration of the violent relationship, and the number of children could also promote understanding of the predisposition/willingness to return to the ex-partner.

5. Conclusions

The current study aims to develop and validate the Attitude Toward Returning to an Ex-partner Scale (ATRES), a self-report instrument designed to assess cognitive and emotional predispositions associated with the tendency to consider reconciliation after an abusive relationship. While not intended as a diagnostic or therapeutic tool per se, the ATRES can support professionals in identifying key beliefs—such as myths about romantic love or internalized messages from the abuser—that may influence decision-making in contexts of gender-based violence or toxic relationships. The scale also contributes to the existing body of research by offering a focused and psychometrically sound instrument that fills a gap in the assessment of factors associated with the willingness to return to a harmful relationship.

The findings support the ATRES as a unidimensional scale with strong internal consistency and relevant content validity. Beyond its psychometric contribution, the scale offers applied value in clinical and psychoeducational settings. In therapeutic contexts, for instance, professionals frequently rely on interviews to explore the motivations and beliefs of individuals who have exited abusive relationships. The ATRES can serve as a complementary tool to help uncover specific cognitive patterns that may not be immediately apparent, providing a structured baseline from which to support reflection and change. At the same time, the instrument remains flexible enough to be used in non-clinical research, particularly in studies investigating factors associated with reconciliation and relationship dynamics following abuse. By targeting a specific construct not addressed by existing measures, the ATRES adds conceptual and practical value to the field of relationship and violence research.