Assessing coercive control in intimate partner relationships: psychometric validation and clinical thresholds of a French version of the Coercive Control Screening Scale (CCSS-F)

ABSTRACT

Background: Coercive control is increasingly conceptualized as a chronic relational pattern of domination that may function as a mechanism of traumatization over time. Sustained exposure to intimidation, isolation, surveillance, and credible threat has been linked to post-traumatic stress disorder (PTSD), complex PTSD (CPTSD), and depressive symptomatology. Despite its relevance for trauma-informed assessment, no validated French-language instrument exists, and no empirically derived cutoff score has been established for the original Coercive Control Screening Scale (CCSS).

Objective: This study aimed to translate and validate a French version of the CCSS (CCSS-F) and to establish the first empirically derived cutoff score to support clinically meaningful screening of coercive control exposure.

Method: Following forward–backward translation and expert panel review, 304 French-speaking women (18–70 years) were recruited, including 225 community participants and 79 women receiving hospital-based psychiatric care for documented intimate partner violence and trauma-related difficulties. Participants completed the CCSS-F and validated measures of battering (WEB), relationship abuse, post-traumatic stress symptoms, depression, relationship dissatisfaction, perceived social support, and social desirability. Factor structure (EFA/CFA), reliability (α, ω), construct validity, and receiver operating characteristic (ROC) analyses were conducted.

Results: The CCSS-F demonstrated excellent internal consistency (α and ω > .90) and a stable unidimensional structure. Strong associations were observed with battering and relationship abuse (ρ > .80), and moderate associations with PTSD and depressive symptoms, supporting convergent validity. Associations with social desirability were non-significant. ROC analyses indicated acceptable discriminatory accuracy (AUC ≈ .78) and supported an empirically derived cutoff score (> 34), representing the first internationally proposed screening threshold for the CCSS.

Conclusions: The CCSS-F is a psychometrically robust and clinically grounded screening instrument. By combining validation in a hospital-based trauma sample with the first empirically derived cutoff score, this study advances trauma-informed identification of coercive control and supports its integration into psychotraumatology research and clinical practice.

HIGHLIGHTS

• First psychometric validation of a French version of the Coercive Control Screening Scale (CCSS-F).

• The CCSS-F showed a robust unidimensional structure and excellent internal consistency.

• Scores were strongly associated with trauma-related symptoms and relationship abuse measures.

• ROC analyses yielded the first empirically derived cutoff score for coercive control screening.

• Findings support integration of coercive control assessment into trauma-informed practice.

1. Introduction

Over the past two decades, coercive control has emerged as a central theoretical and empirical framework for understanding intimate partner violence (IPV), particularly across clinical, legal, and public health research. Moving beyond incident-based definitions that emphasize discrete acts of physical or sexual aggression, coercive control refers to a persistent, intentional, and strategic pattern of domination through which one partner constrains the other's autonomy, freedom, and sense of self, through behaviours such as intimidation, surveillance, isolation, threats, and the micro-regulation of everyday life (Stark, 2007). Rather than constituting a series of isolated abusive events, coercive control tends to be conceptualized as a relational process that structures the victim's environment through chronic threat, constraint, and dependency (Dutton & Goodman, 2005). This process is cumulative and predominantly non-physical, progressively eroding the agency of victims and organizing their daily life around issues of avoidance, appeasement, and survival (Khan & Akram, 2025).

Given its detrimental consequences to mental health and the fact that it is not grounded in highly-visible incidents such as direct physical aggression, it is critical that clinicians and researchers have suitable tools available for the identification of coercive control. This is the focus of the current study.

1.1. Coercive control and trauma

Accumulating empirical evidence indicates that coercive control can be strongly associated with psychological harm – even more so than the frequency or severity of physical violence (Hardesty et al., 2015; Myhill, 2015; Lohmann, Cowlishaw, et al., 2024). On average, women exposed to coercive control report elevated rates of post-traumatic stress disorder (PTSD) (Choudhury et al., 2025), complex PTSD (CPTSD) (Lohmann, Cowlishaw, et al., 2024), depression, anxiety, dissociation, emotional dysregulation (Broughton & Ford-Gilboe, 2017; Crossman et al., 2016; Lohmann, Cowlishaw, et al., 2024), and impaired relational functioning (Hlavaty & Haselschwerdt, 2019). From a psychotraumatology perspective, these findings suggest that coercive control should not be understood merely as a contextual correlate of violence but rather as a chronic relational stressor that can function as a mechanism of traumatization over time (Kassing & Collins, 2026).

Importantly, coercive control challenges narrow, event-based definitions of trauma in IPV. Many coercive tactics – such as isolation, constant monitoring, economic restriction, intimidation, degradation, or implicit threats – do not systematically meet Criterion A for PTSD as defined in the DSM-5-TR (A.P.A., 2023). Nevertheless, these coercive tactics tend to generate an enduring environment of fear, subjugation, and vulnerability. In the ICD-11 framework, which adopts a broader conception of trauma exposure, prolonged experiences of captivity, domination, and credible threat are explicitly recognized as pathways to trauma-related disorders, including Complex-PTSD (C-PTSD).¹ From this perspective, coercive control can be conceptualized as a traumatic process unfolding over time and shaping cognitive schemas, emotional regulation, and relational expectations – rather than as a single traumatic exposure (Kassing & Collins, 2026).

The traumatic impact of coercive control can be further understood through its effects on pre-traumatic, peri-traumatic, and post-traumatic processes. At the pre-traumatic level, coercive control can systematically undermine a victim's fundamental beliefs about safety, justice, autonomy, and interpersonal trust, thereby increasing vulnerability to PTSD and C-PTSD (Lohmann, Cowlishaw, et al., 2024). At the peri- and post-traumatic levels, coercive control shapes threat appraisal, perceived uncontrollability, and anticipatory fear in the victim (Reeves et al., 2025). Recurrent tactics such as isolation, intimidation, deprivation of resources, and implicit or explicit threats create a context in which victims may learn that resistance or attempts to regain autonomy are likely to result in escalation, including physical violence. This climate of credible and ongoing threat is consistent with mechanisms involved in trauma-related psychopathology, including avoidance, emotional numbing, and chronic hyperarousal. Overall, these processes closely align with sound cognitive models of trauma, particularly with regard to maladaptive accommodation of traumatic information, persistent negative self-concept, and disturbances in affect regulation and relationships (Brewin et al., 1996; Cloitre, 2004; Dalgleish, 2004; Ehlers & Clark, 2000).

Despite its profound psychological and traumatic consequences, coercive control remains under-recognized and under-assessed across many clinical and primary care settings. Most IPV screening instruments were developed to identify discrete acts of physical or sexual violence and rely on incident-based approaches that fragment victim experience into isolated events (Arkins et al., 2016; O’Doherty et al., 2015; Rabin et al., 2009). Such approaches are poorly suited to capturing patterns of coercive control unfolding over time, such as chronic domination or relational entrapment. From a trauma-informed clinical perspective, this limitation is critical: failure to identify coercive control may lead clinicians to underestimate ongoing relational threat, misattribute persistent trauma symptoms to individual vulnerability, or conclude prematurely that risk has resolved once physical violence has ceased (Herman, 1997; Myhill & Hohl, 2019). In sum, making assessment tools suitable for coercive control available to researchers and clinicians in particular should be seen as a priority.

1.2. Assessment of coercive control: strengths and limitations of existing tools

A wide range of assessment tools has been used to operationalize coercive control or related constructs, reflecting both the conceptual complexity of the phenomenon and the absence of a standardized measurement approach. Taken together, these tools vary widely in theoretical grounding, length, and intended use. Most do not provide empirically supported thresholds to distinguish clinically meaningful levels of exposure, limiting comparability across studies and constraining their applicability in trauma-focused clinical contexts.

Many studies have relied on broad measures of psychological IPV, such as the Psychological Maltreatment of Women Inventory (Tolman, 1989), or general IPV screening tools including the Women's Experience with Battering (WEB) scale (Smith et al., 1995) and the Woman Abuse Screening Tool (WAST) (Brown et al., 1996). While these tools do capture important aspects of fear, powerlessness and distress in a relationship, they were not designed to assess coercive control as a distinct relational process and do not comprehensively address defining features of coercive control such as surveillance, isolation, and the strategic use of threats to consolidate control.

More specific tools have been developed from theory-driven models of coercive control, including the Coercion in Intimate Partner Relationships (CIPR) scale (Wilson & Fritz, 2023), and the Checklist of Controlling Behaviours (CCB) (Lehmann et al., 2012). These measures offer detailed coverage of coercive tactics and are grounded in clinical and woman-centered theories. Unfortunately, their length and complexity (109 and 84 items respectively) limit their feasibility for routine clinical screening and large-scale trauma-focused research.

A third tool specifically addressing coercive control, the Coercive Control Scale (CCS) (Johnson et al., 2014), represents a major empirical contribution. Derived from population-based survey data, the CCS assesses non-violent controlling behaviours and has demonstrated good internal consistency across samples of current and former partners. The CCS has been instrumental in advancing typological distinctions between forms of IPV, such as intimate terrorism and situational couple violence. However, the CCS was primarily designed for population-level research and behavioural classification, does not provide an empirically derived cutoff score, and was not intended as a trauma-informed screening tool for individual-level clinical assessment.

More recently, Wilson (2022) developed the Coercive Control Screening Scale (CCSS) to address the limitations identified in prior measurement approaches. Drawing explicitly on Dutton and Goodman's (2005) theoretical model, the CCSS was designed as a brief screening instrument capturing survivors’ subjective experience of coercive control, rather than as an exhaustive checklist of coercive tactics. The CCSS includes 11 items and has demonstrated excellent psychometric properties in Canadian community samples (Wilson, 2022), including excellent internal consistency (α = .95), a stable unidimensional structure with excellent fit, and strong concurrent and convergent validity with established measures of battering and psychological abuse (including rs > .70 with physical and emotional victimization in the CCB and rs > .60 with measures of trauma exposure). However, the CCSS is currently unpublished, unavailable in French, and does not come with normative data or a cutoff score, which limits possibilities for clinical interpretation. From a psychotraumatology standpoint, this type of screening scale can be viewed as an aid for identifying individuals exposed to harmful relational dynamics warranting further assessment and intervention; even without constituting a formal diagnostic tool, having normative data along with a cutoff score available is paramount for this type of intended use.

1.3. Rationale for the present study

With coercive control being increasingly recognized as a central driver of psychological harm and complex trauma in IPV (Beck & Raghavan, 2010; Katz et al., 2020; Walklate et al., 2018), a short screening tool with normative data would have many practical uses in psychotraumatology to support early identification of coercive control, inform case formulation and risk assessment, and facilitate the integration of relational threat dynamics into evidence-based trauma treatments. The current study aimed to make such a scale available to French-speaking users by translating, adapting, and providing psychometric validation for the French version of the CCSS (CCSS-F). An additional purpose of this process was to investigate the feasibility of determining a cutoff score to identify the most at-risk women, with implications for use of the CCSS in other languages.

2. Method

The study was conducted in accordance with the Declaration of Helsinki and received approval from the Comité d'Éthique de la Recherche (CEDIS) of Nantes Université (Approval: 26/02/2025; IRB00013074). All participants provided informed consent and were informed of confidentiality procedures, data protection measures, and their right to withdraw at any time. For women in the clinical sample who might have reported distress during assessment, immediate support was available from on-site clinicians.

2.1. The CCSS-F and its translation and adaptation by experts

The scale translation and adaptation were conducted in accordance with established guidelines for cross-cultural adaptation (Beaton et al., 2000), including translation, back-translation, and review by experts and patients. The psychometric properties of the CCSS-F were primarily investigated in a community sample, including its factor structure, reliability based on internal consistency, and validity based on convergent and divergent validity and the scale's ability to discriminate between samples. Data were also collected in a clinical sample, which made it possible to confirm that the scale retained good properties in a clinical group, and to perform a ROC curve analysis to evaluate possible cutoff scores.

The scales chosen to test the convergent and divergent validity of the CCSS-F were as follows. Convergent validity was primarily tested with two measures of IPV (which could be considered a form of concurrent validity, although these two scales do not specifically target coercive control): the Women's Experience with Battering Scale – WEB (Smith et al., 1995) and the Woman Abuse Screening Tool – WAST (Brown et al., 1996). Convergent validity was also tested with measures less specific to IPV: a scale for post-traumatic stress, the International Trauma Questionnaire – ITQ (Cloitre et al., 2018); a scale for depressive symptoms, the Patient Health Questionnaire – PHQ-9 (Kroenke et al., 2001); and a scale for relationship quality expected to correlate negatively with the CCSS-F, the Relationship Assessment Scale – RAS (Hendrick, 1988). Lastly, two measures were used for divergent validity, expected to show little to no correlation with the CCSS-F: a scale for social desirability, the short form of the Marlowe–Crowne Social Desirability (MCSD-C) Scale (Reynolds, 1982), and a scale for perceived social support, the short form of the German Social Support Questionnaire – F-SozU-K6 (Kliem et al., 2015). Women in the community sample completed all scales; women in the clinical sample only completed the WEB to minimize testing time.

The CCSS-F retained the structure of the original scale, as a self-report measure composed of 11 items assessing the frequency of coercive and controlling behaviours experienced within intimate relationships. We used the same 5-point Likert scale, with responses ranging from 0 (Never) to 4 (Very frequently). Total scores were computed as the sum of all items, yielding a total ranging from 0 to 44, with higher scores indicating greater exposure to coercive control. Items for the CCSS-F are available in Table 1.

Table 1.

Items for the CCSS-F.

Item n°	Item	Factor loading on a general factor
Tout d'abord, nous vous demandons de penser à ce que vous ressentez/ressentiez et à ce que vous faites/faisiez en réponse à la personne X pendant la relation:
1	Je me suis senti·e effrayé·e et/ou intimidé·e par la personne X	.78 / .83
2	Je me suis senti·e contrôlé·e et/ou micro-géré·e par la personne X	.72 / .82
3	Je me suis senti·e piégé·e et/ou emprisonné·e dans ma relation avec la personne X	.80 / .78
4	J'ai fait ce que la personne X voulait parce que j'étais intimidé·e et/ou effrayé·e par ce qu'elle ferait si je ne lui obéissais pas (exemples: préjudice physique ou émotionnel)	.85 / .78
5	J'ai fait des choses que je ne voulais pas faire parce que je croyais que la personne X irait au bout de ses menaces (exemples: menaces de me faire du mal, de se faire du mal, ou de nuire à quelqu'un/quelque chose qui m'est cher)	.79 / .78
Maintenant, pensez à ce que la personne X fait/a fait pendant la relation:
6	Pour m’amener à faire ce qu'elle voulait, la personne X a utilisé contre moi des informations personnelles et/ou confidentielles que j'ai partagées avec elle (exemples: erreurs du passé, préférences sexuelles, statut d'immigration)	.74 / .51
7	Pour m'amener à faire ce qu'elle voulait, la personne X a utilisé mes insécurités/vulnérabilités contre moi (exemples: difficultés de santé mentale, faible estime de soi)	.80 / .80
8	La personne X m’a dit ou m’a fait penser que je ne pouvais pas survivre sans elle	.71 / .71
9	La personne X a contrôlé mon accès aux ressources (exemples: argent, transport, nourriture, soins de santé)	.60 / .30
10	La personne X m’a empêché·e ou a rendu difficile pour moi de voir les personnes que j'avais envie de voir (par exemple, amis, famille, médecin, thérapeute)	.72 / .69
11	La personne X gardait la trace de ce que j’ai fait (exemples: consultait mon téléphone, me pistait, demandait à d'autres personnes ce que je faisais, me traquait)	.62 / .58

Note. Factor loadings are for the non-student sample and the student sample, in this order. General instructions were translated as follows: ‘Pour chacune des affirmations ci-dessous, veuillez sélectionner l'option qui explique le mieux à quelle fréquence vous avez vécu les situations suivantes durant votre relation avec la personne X. Si vous répondez aux questions concernant un·e ancien·ne partenaire, veuillez repenser à ce que vous ressentiez durant la relation avec la personne X. Veuillez noter que les exemples ont été utilisés pour clarifier certains points; cependant, les exemples choisis ne saisiront pas l'expérience de tout le monde.’

The CCSS was first translated according to international guidelines for cross-cultural adaptation (Beaton et al., 2000). Two bilingual experts with clinical expertise in psychotraumatology independently translated the scale into French. A third bilingual expert performed a blind back-translation into English. A consensus committee compared all versions, resolving discrepancies through discussion and producing a harmonized version.

Prior to the main validation study, a panel of clinical psychologists and psychiatrists specializing in IPV, trauma and coercive control reviewed the translation of the CCSS to assess its clarity, comprehensibility, and clinical relevance. The experts were first invited to complete the questionnaire as if they were respondents. They then evaluated each instruction and item on two dimensions: clarity (clear / somewhat clear / unclear) and comprehensibility (understandable / somewhat / not understandable). When an instruction or item was rated as partially clear or not clear, experts were encouraged to provide qualitative comments and suggestions for reformulation. The quantitative results of this pre-test showed high overall clarity and comprehension across items, with only a small number of items receiving lower ratings. These elements, along with qualitative feedback, led to minor wording refinements for several items, intended to ensure conceptual equivalence and clinical interpretability in French.

The preliminary version of the scale was then administered to 37 women who were survivors of intimate partner violence and were receiving care in a Psychotrauma Unit. Participants first completed the CCSS-F and then responded to follow-up questions assessing readability, clarity of instructions, and cultural relevance. All items and instructions were rated as clear and understandable by 91–100% of the women (M = 97.6%). No further adjustments to wording were made, and these women were therefore included in the main clinical sample, see Table 1.

2.2. Measures used for convergent and divergent validity

2.3.1. Women's Experience with Battering Scale (WEB)

The WEB (Smith et al., 1995) is a 10-item measure assessing internalized psychological battering, including fear, entrapment, and loss of autonomy within intimate relationships. Items are rated on a 6-point Likert scale, with higher scores indicating greater psychological vulnerability. A French translation, the WEB-F, existed prior to this study (Lamy et al., 2009), but no psychometric validation had been published. Although this was not the core purpose of the present study, we had the same panel of experts involved in developing the CCSS-F also review the WEB items for clarity, comprehensibility, and clinical relevance. The resulting WEB-F version showed a clear one-factor structure (the first four eigenvalues in the community sample: 6.08, 0.75, 0.61, 0.58), and excellent internal consistency in both our community sample (ω_t = .94) and our clinical sample (ω_t = .92). Items for the WEB-F are available in Table 2.

Table 2.

Items and Factor loading for the WEB-F.

Item n°	Item	Factor loading on a general factor
1	Je ressens une insécurité dans ma propre maison	.65 / .45
2	Je me sens honteux·se des choses qu'il/elle me fait	.64 / .65
3	J’essaie de ne pas le/la contredire car j'ai peur de ce qu'il/elle pourrait faire	.73 / .77
4	Je me sens comme programmé·e pour réagir d'une certaine façon à lui/elle	.69 / .77
5	J’ai l’impression qu’il/elle me garde prisonnier·ère	.77 / .74
6	J’ai l’impression de ne plus avoir le contrôle sur ma propre vie, ni de pouvoir, ni de protection	.81 / .79
7	Je cache la vérité aux autres car j'ai peur de la dire	.67 / .68
8	J’ai l’impression que je lui appartiens et qu'il/elle me contrôle	.76 / .73
9	Il/elle peut m’effrayer sans lever la main sur moi	.81 / .67
10	Il/elle a un regard qui me transperce et me terrifie	.74 / .63

Note. Factor loadings are for the non-student sample and the student sample, in this order. General instructions were translated as follows: ‘Voici des affirmations que des personnes ont utilisées pour décrire leur vie avec leur partenaire violent. Veuillez lire chaque affirmation, puis sélectionner la réponse qui correspond le mieux à votre degré d'accord ou de désaccord général. Il n'y a pas de bonnes ou de mauvaises réponses.’

2.3.2. Woman Abuse Screening Tool (WAST)

The WAST is an eight-item screening measure assessing tension, conflict, and indicators of psychological or physical abuse within intimate relationships (Brown et al., 1996). Items are rated on a 3-point scale (0 = No tension, 2 = A lot of tension), with higher scores reflecting greater relational strain. The original English version showed excellent reliability and a strong single-factor structure, and has been widely used in clinical and primary-care settings for IPV screening. A French version of the WAST (WAST-F) has also demonstrated very good acceptability and diagnostic performance in prior research (Guiguet-Auclair et al., 2021). Internal consistency was good in our community sample (ω_t = .90), and the scale showed a clear unidimensional structure, with all items loading adequately on a single factor.

2.3.3. International Trauma Questionnaire (ITQ)

The ITQ is an 18-item measure assessing PTSD and CPTSD based on ICD-11 criteria (Cloitre et al., 2018). Six items assess PTSD symptoms (re-experiencing, avoidance, sense of threat) and six items assess Disturbances in Self-Organization (DSO: affective dysregulation, negative self-concept, relational difficulties). Each symptom cluster is accompanied by three items assessing functional impairment. Items are rated on a 5-point scale (0 = Not at all, 4 = Extremely), referring to the past month. The original ITQ has demonstrated excellent psychometric properties, and the French adaptation (ITQ-F) similarly confirmed the ICD-11 two-factor structure and strong reliability (Peraud et al., 2022). Internal consistency in our community sample was excellent (ω_t = .92 for the PTSD subscale and ω_t = .88 for the DSO subscale), with sufficient variability to allow for meaningful data analysis according to mean symptom scores (for PTSD: M = 13.85 out of 36, SD = 10.00; for DSO: M = 14.29 out of 36, SD = 8.32).

2.3.4. Patient Health Questionnaire (PHQ-9)

The PHQ-9 (Kroenke et al., 2001) is a nine-item measure assessing the frequency of depressive symptoms over the past two weeks. Items are rated on a 4-point scale (0 = Not at all to 3 = Nearly every day), with total scores ranging from 0 to 27. The PHQ-9 shows strong psychometric properties in its original validation and in French-language samples (Arthurs et al., 2012), including large Québec community studies (Gauvin et al., 2024). Internal consistency in our community sample was excellent (ω_t = .93), with sufficient variability according to mean symptom scores (M = 1.12, SD = 0.77).

2.3.5. Relationship Assessment Scale (RAS)

The RAS (Hendrick, 1988) assesses satisfaction with the quality of a relationship; a French version is available under the name Échelle d'Évaluation de la Relation (EER; Saramago et al., 2021). The scale comprises seven items rated on a 5-point Likert scale (1 = low satisfaction to 5 = high satisfaction). The French EER has demonstrated a unidimensional structure and good psychometric properties including measurement invariance across genders; internal consistency was also excellent in our community sample with ω_t = .92.

2.3.6. Perceived Social Support Scale (F-SozU-K6)

The German Social Support Questionnaire (F-SozU-K6) is a six-item measure of general perceived social support (Kliem et al., 2015). Items are rated on a 5-point Likert scale (1 = Not at all true to 5 = Very true), with higher scores reflecting higher levels of perceived social support. The F-SozU-K6 is widely used and has shown excellent psychometric properties in two large representative German population samples, demonstrating excellent reliability and strong evidence for strict unidimensional factorial invariance across age and gender. The F-SozU-K6 showed excellent internal consistency in our community sample (ω_t = .94). An exploratory factor analysis further supported the expected unidimensional structure, with all items loading strongly on a single factor.

2.3.7. Marlowe–Crowne Social Desirability Scale – Short Form C (MCSD-C)

The MCSD-C (Reynolds, 1982) assesses social desirability with 13 items rated in a true/false format. Higher scores indicate a stronger tendency to provide socially desirable responses. The short form has demonstrated acceptable reliability and strong convergence with the original 33-item scale (Reynolds, 1982). No French-language version of the scale has undergone formal psychometric validation, although one translation has been used in a large cross-cultural study including French-speaking samples from Switzerland and several African countries, showing adequate reliability and a unidimensional structure (Verardi et al., 2010). We translated the scale for the current study, using a forward–backward translation procedure with expert review. Internal consistency was low in our community sample (ω_t = .61). The scale was kept for complete reporting, although this low reliability could limit relations with other measures.

2.3. Participants and procedure

2.3.1. Community sample

The community sample consisted of 225 French-speaking women recruited online between March and September 2025. Inclusion criteria required participants to (1) self-identify as a woman, (2) be aged 18 years or older, (3) be a native French speaker, and (4) have been involved in at least one intimate relationship of a minimum duration of three months where they experienced relational difficulties. Participants who did not provide complete data at least for the CCSS were excluded from the sample.

Participants were recruited from university mailing lists, social media advertisements, and women's community networks. Due to this recruitment procedure, the sample mostly included participants with one of two profiles: young students, and non-students aged 30 and above. To avoid potential issues with the analyses (given the higher risk of having experienced significant IPV in older participants) and as a way to cross-validate the robustness of psychometric properties of the CCSS-F, we split the sample into two sub-samples and examined them separately. The sample of students comprised 104 participants (mean age = 21.57 years, SD = 2.31, range = 18–29); the sample of non-students comprised 119 participants (mean age = 44.17 years, SD = 10.79, range = 22–78).

The data were collected online on a secure Qualtrics platform. Participants completed the EER, CCSS-F, WEB, WAST, ITQ, PHQ-9, F-SozU-K6, and MCSD-C, in this order. Participants were instructed to complete the scales thinking of any past or current intimate relationship, of a minimum duration of three months, where they experienced relational difficulties. Items within a scale could not be left blank (except for items in the ITQ, which allows partial subscale completion). Average completion time was approximately 20–25 min. Participation was anonymous and participants could withdraw at any time without justification. No compensation was offered.

2.3.2. Clinical sample

The clinical sample comprised 79 women (mean age = 43.90 years, SD = 11.25 years, range = 23–72 years). Participants in this group had a mix of study levels: below high school (n = 28), end of high school (n = 20), or college education (n = 31). Most participants were single at the time of testing (n = 66).

Women were recruited between May and July 2025 in partner services specialized in psychotrauma care. All participants in this sample were actively seeking treatment for IPV-related psychotrauma symptoms and were recruited during routine consultations in specialized psychotrauma services.

Clinicians informed eligible women about the study during routine consultations. Inclusion criteria were to (1) self-identify as a woman, (2) be aged 18 years or older, (3) be a native French speaker, and (4) currently receive care for psychotrauma or interpersonal violence–related symptoms due to IPV (Battered Person Syndrome, 6B40 in the ICD-11). Exclusion criteria were acute psychiatric instability or cognitive impairment preventing reliable consent.

The data were collected in the course of routine care as part of a structured clinical assessment. After providing written informed consent, participants completed the CCSS-F then the WEB. Participants were instructed to complete both scales thinking of the abusive relationship leading to IPV for which they were currently receiving treatment. The scales were completed in paper format within the clinical setting. Clinical staff were available to provide clarification but did not intervene in participants’ responses. Participation had no impact on treatment decisions and the data were fully anonymized before analysis. All participants in this sample provided complete data for all scales.

2.4. Data analysis

Reliability was estimated based on internal consistency, as assessed using McDonald's total omega coefficient (ω_t), a more robust alternative to Cronbach's alpha coefficient (Zinbarg et al., 2005); computed using the psych package for R (Revelle, 2025). Internal consistency was estimated separately for the two community sub-samples (students vs. non-students) and for the clinical group. Convergent and divergent validity were assessed with bivariate Pearson correlations between the CCSS-F and all other scales (the WEB in the clinical sample, and the WEB, WAST, ITQ, PHQ-9, RAS, F-SozU-K6 and MCSD-C in the two community sub-samples).

Known-groups validity was tested by comparing average CCSS-F scores between the community sample and the clinical sample; to avoid any difference being driven by socio-demographic differences, we first downsampled the community sample to create a subgroup matched to the clinical sample. This was done using propensity score matching (with the matchit package for R Ho et al., 2011) to select an optimal match in terms of age and study level within the community sample, for each participant in the clinical sample. This procedure yielded two matched groups of the same size (i.e. n = 79 for both the clinical group and the matched community subgroup). The two matched groups no longer differed on age (p = .816), but there were residual differences in educational attainment due to the larger proportion of women in the clinical group with education below high school (p < .001). For this reason, the difference between the two groups was tested using an analysis of covariance (ANCOVA) controlling for study level.

Structural validity was only investigated in the larger community sample. It was primarily examined with confirmatory factor analysis (CFA), conducted using Diagonally Weighted Least Squares (DWLS) as appropriate for ordinal data (estimated using the lavaan package for R (Rosseel, 2012)). Model fit was evaluated using scaled (not robust) χ², CFI, TLI, RMSEA, and SRMR, interpreted without strict cutoffs (McNeish & Wolf, 2023; Palíšek et al., 2025). This CFA was supplemented with exploratory factor analysis (EFA) to search for alternative meaningful factors (principal axis factoring with oblimin rotation). EFA suitability was tested with Bartlett's test of sphericity and the Kaiser–Meyer–Olkin index. The number of factors was decided based on a combination of Cattell's scree test and parallel analysis.

To investigate the feasibility of determining an adequate cutoff score, a ROC curve analysis was conducted by comparing the clinical sample with its matched community subgroup. The possibility of adequately discriminating between the two groups was quantified using the Area Under the Curve (AUC). Sensitivity, specificity, and classification accuracy were computed for all possible cutoffs; the optimal cutoff was determined using Youden's index (J = Sensitivity + Specificity – 1).

3. Results

3.1. Descriptive statistics and discriminating power

Descriptive statistics for all measures in the community sample are detailed in Tables 3–5. For the two community samples, participant disengagement due to the survey length meant there were less data available for some scales (with a minimum of n = 89 in the non-student sample and n = 56 in the student sub-sample for the last scale, the MCSD-C), but all scales had enough data available for a reasonable test of convergent and divergent validity.

Table 3.

Descriptive statistics for all measures in the community sample of non-students.

Measure	Community sample (non-students)
	n	M	ET	Skewness	Kurtosis	Range
CCSS-F	119	24.63	11.58	−0.18	−0.93	1–44
WEB-F	112	40.42	13.60	−0.46	−0.65	10–60
WAST	110	9.01	3.44	−0.07	−0.24	0–16
ITQ (Total)	97	29.00	18.20	0.21	−1.02	1–68
ITQ (PTSD)	103	14.91	10.67	0.24	−1.09	0–36
ITQ (DSO)	97	14.01	8.53	0.15	−0.93	0–32
PHQ-9	86	11.50	7.72	0.53	−0.43	0–30
EER	119	16.50	5.74	0.78	0.35	7–34
F-SozU-K6	91	12.78	11.04	0.33	−1.41	0–30
MCSD-C	89	18.89	2.30	0.31	−0.16	14–25

Note. The possible range of scores was 0–44 for the CCSS-F and 10–60 for the WEB-F. Internal consistency for all scales used for convergent and divergent validity is detailed in the Methods section.

Table 4.

Descriptive statistics for all measures in the community sample of students.

Measure	Community sample (students)
	n	M	ET	Skewness	Kurtosis	Range
CCSS-F	104	12.02	10.06	0.71	−0.48	0–37
WEB-F	92	26.74	12.63	0.49	−1.04	10–51
WAST	90	5.50	3.49	0.54	−0.32	0–15
ITQ (Total)	64	26.86	14.72	0.37	−0.45	1–65
ITQ (PTSD)	65	12.15	8.66	0.53	−0.57	0–33
ITQ (DSO)	64	14.72	8.02	0.26	−0.63	0–34
PHQ-9	55	12.22	6.51	0.32	−0.29	0–27
EER	104	20.29	6.48	0.33	−0.80	8–34
F-SozU-K6	56	11.84	9.28	0.24	−1.29	0–30
MCSD-C	56	19.63	2.15	0.10	−0.56	15–24

Note. The possible range of scores was 0–44 for the CCSS-F and 10–60 for the WEB-F. Internal consistency for all scales used for convergent and divergent validity is detailed in the Methods section.

Table 5.

Descriptive statistics for all measures in the clinical sample.

Measure	Clinical sample
	n	M	ET	Skewness	Kurtosis	Range
CCSS-F	79	34.22	8.57	−1.09	0.97	6–44
WEB-F	79	52.62	9.08	−2.01	4.93	15–60

Note. The possible range of scores was 0–44 for the CCSS-F and 10–60 for the WEB-F.

The distribution of total CCSS-F scores is depicted in Figure 1. Overall, scores on the scale covered the whole range of possible values. The community sample of students displayed positive asymmetry, consistent with generally lower experience of IPV, whereas the clinical sample displayed negative asymmetry, consistent with very high experience of IPV, and the community sample of non-students showed a balanced distribution.

Figure 1.

Distribution of total CCSS-F scores for the community student sample, the community non-student sample, and the clinical sample.

Descriptive statistics (Tables 3–5) confirmed that the skewness and kurtosis of the scale were reasonably close to normal in all three groups, indicating that the scale retained adequate discriminating power in all samples without a major ceiling effect or floor effect even in the student and clinical samples. Item-level inspection of descriptive statistics indicated reasonable distributional properties for all items, with all response options for all items being selected by at least a minimal number of participants.

3.2. Reliability (internal consistency) of the CCSS-F

The total score of the CCSS-F demonstrated excellent reliability for the community sub-sample of students (ω_t = .93), the community sub-sample of non-students (ω_t = .95), and the clinical sample (ω_t = .91). No item of the CCSS-F increased internal consistency when removed, in any of the samples.

3.3. Factor structure of the CCSS-F

CFA demonstrated that a factor structure with a single general factor for the CCSS-F fit relatively well in the community sample, for both the non-student sub-sample (scaled fit: χ²(44) = 102.43, p < .001, CFI = .98, TLI = .98, RMSEA = .106, SRMR = .061; all loadings ≥ .68) and the student sub-sample (scaled fit: χ²(44) = 62.83, p = .033, CFI = .99, TLI = .99, RMSEA = .064, SRMR = .075; all loadings ≥ .53). Slightly elevated values for the RMSEA and SRMR were driven by residual covariances between items with similar contents, but these were not consistent between the two sub-samples, suggesting that there was no stable more complex structure to model.

To confirm that the CCSS-F can be considered reasonably unidimensional, an additional EFA was conducted (significant Bartlett test with ps < .001 in both sub-samples; KMO = .921 for non-students and .907 for students, confirming factorability). Parallel analysis and scree plot inspection both supported a clear single-factor solution (first eigenvalues in the non-student sample = 6.51, 1.03, 0.64, 0.55; first eigenvalues in the student sample = 5.93, 1.04, 0.88, 0.70), with the first factor explaining 50–55% of total variance (all factor loadings ≥ .30). In sum, the CCSS-F demonstrates a predominantly unidimensional structure across samples: although the scale seems to include subsets of items that are not quite interchangeable, it supports computation of a single total score.

3.4. Known-groups validity

As expected, women in the clinical group reported substantially higher CCSS-F scores (M = 34.22, ET = 8.57) than their matched controls from the community sample (M = 22.46, ET = 12.06). The difference between the two groups was significant with a very large effect size, even when controlling for study level, F(1, 155) = 48.16, p < .001, ${\eta }_p^2$ = .24. Using a non-parametric test to account for skewness in the clinical group led to the same conclusion (Mann–Whitney U = 1368, p < .001).

3.5. Concurrent, convergent, and divergent validity

Bivariate correlations between all measures in the community samples of students and non-students are summarized in Table 6.

Table 6.

Bivariate correlations between all measures in the two community samples.

Variable	1.	2.	3.	4.	5	6.	7.	8	9
1. CCSS-F	–	.84	.83	.14	.24	.06	−.60	−.02	−.02
2. WEB-F	.85	–	.80	.29	.32	.02	−.67	−.10	.13
3. WAST	.77	.75	–	.16	.29	−.04	−.70	.07	.08
4. ITQ (PTSD)	.67	.59	.56	–	.55	.57	−.10	−.23	.15
5 ITQ (DSO)	.52	.48	.39	.76	–	.59	−.34	−.31	.27
6. PHQ-9	.57	.50	.47	.75	.78	–	.09	−.20	.22
7. EER	−.63	−.59	−.63	−.43	−.28	−.39	–	.04	−.19
8. F-SozU-K6	−.24	−.32	−.27	−.43	_.51	−.44	.20	–	−.04
9. MCSD-C	−.05	.08	−.10	.06	.22	.09	.02	−.18	–

Note. Correlations above the diagonal are for the community sample of students; correlations below the diagonal are for the community sample of non-students. Significant correlations at the p < .05 level are in boldface.

Overall, the CCSS-F had excellent convergent validity with measures of relationship abuse and battering, with all correlations with the WEB and WAST being ≥ .77 in the two community sub-samples. Convergent validity was also excellent with relationship quality, with negative correlations ≥ −.60 in the two sub-samples. Convergent validity with post-traumatic stress and depression was very good in the community sample of non-students, with correlations with the ITQ PTSD and DSO subscales and the PHQ-9 all ≥ .52, but the same correlations were non-significant in the community sample of students, presumably due to the low prevalence of IPV in this sample. Lastly, divergent validity with perceived social support and social desirability was very good, with all correlations with the F-SozU-K6 and MCSD-C being very weak and only the correlation between CCSS-F and F-SozU-K6 reaching significance in the community sample of non-students, r = −.24, p = .022. Of secondary interest, the translated WEB-F also showed excellent validity, with correlations with other measures very similar to the CCSS-F. In the clinical sample, the CCSS-F showed a very high correlation with the WEB-F, r = .61, p < .001, supporting convergent validity even in participants with major IPV.

3.6. ROC analysis and clinical cutoff determination

The final step was to determine possible cutoff scores by comparing the clinical sample and matched controls. This series of analyses should be interpreted with the distribution of observed scores in mind (see Figure 1): the substantial overlap between the community sample of non-students and the clinical sample is a reminder that women in the community sample, although serving as controls for this analysis, may also experience significant levels of IPV; in other words, women in the control group whose score is above a given cutoff may not be ‘false positives’ in the traditional sense.

The ROC curve indicated acceptable discriminatory ability, with an area under the curve AUC = .78, consistent with a moderate-to-good capacity to distinguish women receiving dedicated psychotrauma care for IPV from women in the general population. The ROC curve is displayed in Figure 2. All possible cutoffs for total CCSS-F scores (between 0 and 44) were examined, and for each threshold the corresponding sensitivity and specificity were computed (see Table 7). For reference, a woman scoring on the middle point for all items of the CCSS-F would obtain a score of 22.

Figure 2.

ROC curve for the discrimination between the clinical sample and matched controls for the community sample. AUC = .78.

Table 7.

Sensitivity and specificity for all possible CCSS-F score cutoffs.

Possible cutoff for the total CCSS-F score	Sensitivity (% clinical sample above the cutoff)	Specificity (% matched controls below the cutoff)	Suggested cutoff label
15.5	95%	33%
16.5	95%	35%
17.5	95%	38%
18.5	94%	38%
19.5	94%	42%
20.5	94%	46%
21.5	94%	47%
22.5	92%	49%
23.5	90%	49%	Moderate risk
24.5	86%	52%
25.5	84%	53%
26.5	81%	56%
27.5	77%	63%
28.5	75%	63%
29.5	73%	66%	High risk
30.5	68%	72%
31.5	67%	75%
32.5	67%	77%
33.5	66%	78%
34.5	65%	82%	Very high risk
35.5	58%	84%
36.5	47%	85%
37.5	42%	89%
38.5	35%	90%	Extreme risk
39.5	30%	92%
40.5	27%	94%
41.5	19%	95%
42.5	18%	97%
43.5	14%	97%

Note. Data not shown for cutoffs 0.5–14.5. Suggested cutoffs are in boldface.

The Youden index identified a CCSS-F score of 34.5 as the optimal compromise, yielding a sensitivity of 64.5% and a specificity of 82.3%. In other words, approximately two-thirds of clinical cases had a total CCSS-F score of 35 or above and were correctly classified at risk of coercive violence, but about one-fifth of the control group also had a score 35 or above. Given the trade-offs inherent in single-threshold decisions, and the risk of women in the control sample actually suffering from significant levels of IPV, we suggest the use of additional cutoffs to provide clinicians with graded indicators of severity or risk:

• 23.5 – moderate risk: sensitivity 90%, specificity 49%

• 29.5 – high risk: sensitivity 73%, specificity 66%

• 34.5 – very high risk: sensitivity 65%, specificity 82% (Youden-optimal)

• 38.5 – extreme risk: sensitivity 35%, specificity 90%.

4. Discussion

4.1. Summary of the main findings

This study provides psychometric validation for a French version of the Coercive Control Screening Scale (CCSS-F), as well as the first empirically derived cutoff score for this tool. The results indicate that the CCSS-F has adequate discriminating power, a clear unidimensional structure supporting the use of a single total score, excellent internal consistency, and excellent validity including concurrent/convergent, divergent, and known-groups validity. Critically, the CCSS-F scores adequately discriminated between participants in a community sample and in a clinical sample exposed to IPV, and it was possible to derive cutoff scores with adequate sensitivity and specificity. Together, these findings support both the use of the CCSS-F as a brief, psychometrically sound screening tool that addresses a critical gap in trauma-informed assessment within French-speaking contexts, and more generally the use of the CCSS (Wilson, 2022) as a scale capable of capturing clinically significant exposure to coercive control – also suitable for translation to other languages.

4.2. Contribution to the assessment literature and to trauma-informed practice

The current study contributes to ongoing efforts to improve the assessment of coercive control by addressing several well-documented limitations in the literature. Previous research has relied heavily on broad measures of psychological IPV, author-created item sets without norming or cutoff scores, or lengthy instruments listing specific behaviours. This resulted in substantial heterogeneity, limited comparability across studies, and difficulty in translating the methods to actual use in clinical contexts (Hamberger et al., 2017; Khan & Akram, 2025; Myhill & Hohl, 2019). The CCSS-F offers a parsimonious alternative that prioritizes survivors’ lived experience of coercive control over exhaustive enumeration of behaviours, thereby aligning measurement more closely with theoretical conceptualizations of coercive control as a subjective relational process – in the context of caring for psychological trauma.

The establishment of an empirically derived cutoff score represents a further methodological advance. A scale without a clear basis for decision-making has limited practical utility for clinical purposes; this applies to existing coercive control measures in general, including well-established instruments such as the Coercive Control Scale developed by Johnson and colleagues (2014). In the present study, the feasibility of determining a cutoff based on ROC analyses confirms that the CCSS can be suitable as a screening tool in actual clinical use. As the CCSS was translated as closely as possible rather than adapted, the same conclusion generalizes to the use of the CCSS in other languages, although investigators operating in different cultural contexts may want to derive alternate cutoff scores more relevant to samples exposed to less or more coercive violence.

It is necessary to reiterate here that cutoff scores inevitably simplify complex phenomena. Such pragmatic thresholds are commonly used in trauma and mental health research to facilitate early identification and to guide further assessment, provided their limitations are explicitly acknowledged (Šimundić, 2009). The notable amount of overlap between community and clinical samples (Figure 1) is a reminder that elevated levels of coercive control can be found in relationships outside of clinical contexts, which makes it impossible to clearly delineate a ‘problematic’ level of coercive control. Similar results were obtained in the original validation samples for the CCSS (Wilson, 2022). In other words, a woman scoring below the cutoff that best separates a non-clinical group from a clinical group treated for IPV can still be exposed to concerning levels of coercive control in a relationship. For this reason, the cutoff scores provided with the CCSS-F should be understood as supporting screening and clinical orientation, rather than diagnostic classification. Likewise, the CCSS-F should not be used in isolation but as part of a broader clinical assessment that considers safety, comorbidity, and contextual factors.

With these caveats in mind, the CCSS-F as a brief screening tool can significantly help identify women experiencing coercive control in situations where physical violence is absent, minimized, or no longer ongoing; in such contexts trauma-related symptoms may otherwise appear unexplained or resistant to treatment. Incorporating assessment of coercive control into routine trauma evaluation may enhance case formulation, improve risk assessment, and support more tailored interventions. For trauma-focused clinicians, including those working with cognitive–behavioural, EMDR, or integrative approaches, recognizing coercive control as an ongoing or past relational threat may be critical for understanding symptom persistence, relational avoidance, or difficulties in emotional regulation (Hameed et al., 2020). Used appropriately, the CCSS may contribute to bridging the gap between IPV research and psychotraumatology by making coercive control more visible within trauma-focused care.

4.3. Coercive control as a chronic traumatic process

Beyond psychometric validation, the current results – and more specifically the strong associations between the CCSS-F and other measures (Table 6) – have important implications for how coercive control is conceptualized within psychotraumatology.

Overall, the excellent validity of the CCSS-F, as a scale designed to assess overall subjective experience, confirms that coercive control does not have to be understood as a discrete traumatic exposure to a series of behaviours. The strong observed associations between CCSS-F scores and both PTSD and DSO symptoms, as well as depressive symptoms, in a non-student sample, in particular, are consistent with a growing body of evidence indicating that the subjective experience of being exposed to coercive control is strongly linked to psychological trauma, often independently of physical violence (Crossman et al., 2016; Atig Le Griguer & Boudoukha, 2025; Kassing & Collins, 2026; Lohmann, Felmingham, et al., 2024). The strong negative association observed between coercive control and relationship satisfaction further suggests that the CCSS-F captures not only exposure to controlling behaviours but also the subjective experience of entrapment and powerlessness that is central to trauma-related harm.

Taken together, these results confirm that coercive control can be approached not as a series of discrete behaviours, but as a chronic relational process of cumulative and insidious traumatization. Repeated exposure to credible threat, enforced dependence and loss of agency, and domination and constraint can progressively shape trauma-related symptomatology. This aligns with models of cumulative trauma and chronic stress that emphasize the pathogenic role of prolonged exposure to uncontrollable stressors rather than singular catastrophic events (Briere et al., 2016; Eissa & Zeira, 2024; Herman, 2015). In this sense, coercive control may help explain why some individuals develop severe and persistent trauma-related symptoms in the absence of events that clearly meet Criterion A for PTSD under the DSM-5-TR framework (A.P.A., 2023), while being fully consistent with the broader trauma definition adopted in ICD-11 (WHO, 2022). In turn, this is a reminder that risk assessments focused purely on observable behaviours risk overlooking situations where women are exposed to elevated levels of IPV and risk significant psychological harm.

4.4. Coercive control in the broader context of PTSD research

Although a relatively recent addition to the literature in psychotraumatology, coercive control, as a construct, fits rather well with classic models of PTSD. Coercive control likely exerts its traumatic impact through multiple, interacting mechanisms, as reflected in the nature of the items composing the CCSS. At the pre-traumatic level, CCSS items target a sustained domination process systematically undermining fundamental beliefs about safety, autonomy, justice, and interpersonal trust. At the peri- and post-traumatic levels, CCSS items are oriented towards continuous surveillance, intimidation, and implicit or explicit threats, fostering a state of chronic hypervigilance and anticipatory fear consistent with core PTSD symptoms such as hyperarousal and avoidance. All these disruptions map closely onto cognitive models of PTSD and complex PTSD, which emphasize maladaptive accommodation of traumatic information, persistent negative self-concept, and disturbances in affect regulation and relationships (Brewin & Holmes, 2003; Cloitre, 2004; Ehlers & Clark, 2000; Ehring et al., 2006).

Indirectly, the present findings are particularly relevant for ongoing debates regarding the distinction between PTSD and complex PTSD. Coercive control shares several defining features with exposure profiles typically associated with CPTSD, including chronicity, interpersonal threat, and constrained escape (Karatzias et al., 2017; Maercker et al., 2022). Although the present study was not designed to test differential diagnostic pathways, the separate analyses of the ITQ dimensions showed that CCSS-F scores were associated not only with PTSD symptoms but also with DSO symptoms in the non-student community sample. This pattern of associations observed suggests that coercive control may be especially relevant for understanding disturbances in self-organization, such as affect dysregulation, negative self-concept, and relational difficulties. In this respect, the CCSS-F may provide a useful complementary tool for identifying relational trauma contexts that are not adequately captured by traditional trauma exposure checklists.

4.5. Directions for future studies

Although our focus was on providing psychometric validation for the CCSS-F and assessing the feasibility of deriving a cutoff score (more than testing causal pathways), our results contribute to the broader literature providing evidence for strong associations between coercive control and relevant outcomes, such as PTSD or depression symptoms or relationship satisfaction, with a few limitations. First, our design was purely cross-sectional; studies interested in confirming the associations between coercive control and other measures may be interested in implementing longitudinal designs to examine how coercive control trajectories interact with symptom development, maintenance, and recovery. Second, we used exclusively self-report measures; these measures are highly relevant in a context where coercive control is approached as a subjective experience (and the lack of association with social desirability suggests limited systematic distortion), but other-report measures (e.g. symptom assessment by a psychologist) would help confirm that the current associations between coercive control and other measures were not overestimated by recall or response biases. Third, although we examined PTSD and DSO dimensions of the ITQ separately, the present study did not include a formal diagnostic assessment of PTSD or CPTSD. Future longitudinal and clinical studies integrating the CCSS-F with dimensional assessments of PTSD and CPTSD would help refine understanding of how relational trauma processes contribute to distinct trauma symptom profiles over time.

The CCSS-F itself demonstrated excellent psychometric properties overall, with the exception of an imperfect fit for a unidimensional factor structure in CFA (with residual covariances between items not entirely accounted by a general factor). The EFA clearly supported a one-factor solution, so this result does not jeopardize the use of a CCSS-F score for clinical decision-making, but it does invite further discussion. Similar patterns have been documented in other trauma-related screening tools, where strict CFA criteria may be difficult to meet despite strong reliability and validity evidence (Marsh et al., 2004; Reise et al., 2013), so this suboptimal fit for a unidimensional structure may just reflect the inherent challenge of modelling complex relational constructs using brief self-report instruments, but other interpretations are possible. Methodologically, residual covariances between items might partly reflect the fact that the scale was not primarily designed with unidimensionality in mind, as should be clear from its structure (the CCSS comprises two sub-parts with different instructions, and some of the item phrasings partly overlap; see Table 1). Theoretically, this imperfect unidimensionality might also reflect the complex processes at play in coercive control; better understanding of the pathways by which coercive control shapes individual trajectories could help refine the dimensionality of future assessments.

Another possible direction for future studies is placing greater focus on participants other than women: the current study exclusively included women, which obviously limits generalizability to other populations. This choice was consistent with the gendered patterns of coercive control documented in the literature, with the study's clinical objectives, and with recruitment constraints (men were allowed to participate in the study, but there were so few non-women responses that only case study analyses would have been possible). However, coercive control can also be observed in individuals other than women, and it is unknown whether the CCSS-F or its cutoff would have adequate properties in other genders.

Lastly and more generally, the results obtained with the CCSS appear to strongly depend on the type of sample and instructions. The current study asked participants to report on a current or past relationship where they experienced relational difficulties; non-student women (aged about 44 years on average) had elevated CCSS-F scores on average (mean = 24.63, median = 25 on a scale from 0 to 44) and showed a significant relation with PTSD and depression symptoms; student women (aged about 21 years on average) had much lower CCSS-F scores (mean = 12.02, median = 9.5 on a scale from 0 to 44) and showed no relation with PTSD and depression symptoms. By contrast, the original validation sample in English-speaking women (Wilson, 2022, Study 4) included mostly non-student women (aged about 47 years on average), asked participants to report on any current or past relationship, and found CCSS-F scores much lower and closer to our student sample (median = 19 on a scale from 11–55 corresponding to a median = 8 on our scale from 0 to 44).

These differences between samples could be driven by a mix of differential exposure to coercive violence with age, participants choosing different relationships to report on, and/or cross-cultural differences between countries in relationship dynamics. In turn, this means that investigators using the CCSS or CCSS-F have to carefully consider instructions regarding which relationship participants are asked to assess, and to apply cutoff scores to other samples only with care. In other words, the cutoff scores suggested here (Table 7) are primarily valid for middle-aged French women (corresponding to the composition of our clinical group and matched controls) reporting on a difficult relationship. These cutoffs are relevant for typical clinical uses, and might also work well in similar samples also reporting on a difficult relationship (e.g. 19.5% of student women reporting on a difficult relationship in the current sample were above the cutoff for moderate risk of coercive control), but they could be completely unsuitable in other contexts. In turn, this calls for replication in larger and more diverse samples, and for cross-cultural comparison studies examining measurement invariance and the predictive validity of coercive control across different linguistic and cultural contexts.

5. Conclusion

This study provides strong evidence that the CCSS-F (and WEB-F) has excellent validity for assessing coercive control among French-speaking women. By demonstrating the feasibility of using an empirically derived cutoff score to identify coercive control in the context of IPV, these results also advance trauma-informed assessment and address a critical gap in the psychotrauma literature. More broadly, the results highlight the importance of conceptualizing coercive control as a relational process with significant traumatic consequences, and underscore the need to integrate the assessment of coercive control into trauma-focused research and clinical practice.

Funding Statement

This research received institutional support from Nantes Université and the Laboratoire de Psychologie des Pays de la Loire (LPPL, UR 4638). No external funding was obtained for this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethical approval and consent to participate

The study was approved by the local ethics committee from Nantes University (CEDIS 26/02/2025, approval n°IRB00013074). All participants provided informed consent prior to participation. The study complied with the Declaration of Helsinki.

Consent for publication

Not applicable, as no identifying information is included.

Availability of data and materials

Data supporting the findings of this study are available from the corresponding author upon reasonable request and subject to ethical restrictions.