Design and Validation of the Moral Courage in Bullying Scale in Adolescents

Moral Courage in Bullying Bystander Behaviors

Courage is understood as the capacity to act despite the presence of fear, risk, or threat (Pury & Lopez, 2010). It entails a cognitive-moral process through which individuals translate moral intention into real action (Baumert et al., 2024). Moral courage, unlike other forms of prosocial behaviors aimed at helping others, is triggered when bystanders perceive a situation as risky and lacking personal benefit yet still choose to intervene (Niesta et al., 2010). Although some studies question the prosocial nature of moral courage due to its inherently confrontational nature (Sasse et al., 2022b), recent research suggests that it is more closely associated with an ethics of care and cooperative forms of prosocial behavior than with aggressive and impulsive behaviors (García-Carrera et al., 2026).

The theoretical foundations of moral courage draw on multiple perspectives, reflecting the evolution of the construct. For Rest’s (1986) Four Component Model, it is framed as part of moral character, being the inner strength that enables people to translate moral judgment into action, where emotion and motivation, play a central role in sustaining this process despite risks. Bandura’s (1991) social cognitive theory of moral agency highlights the importance of resisting moral disengagement and sustaining a sense of moral self-efficacy to act courageously under adverse conditions. Virtue-based approaches conceptualize it as a strength of character grounded in integrity and coherence between values and behavior (Pury & Lopez, 2010). Integrative socio-psychological frameworks (Li et al., 2025; Osswald et al., 2010) suggest that moral courage emerges from the convergence of moral judgment, emotional-motivational processes and contextual influences, enabling individuals to translate moral judgment into action within socially risky situations.

Moral courage has been identified with other-oriented emotions, like moral anger or sympathy, which are felt when individuals show a personal disposition to help others (Malti et al., 2021), but it is also opposed to other moral emotions, such as schadenfreude, which is not associated with concern for others’ well-being. Experimental studies highlight moral anger as a stronger predictor of intervention than other emotions, such as guilt or fear (Sasse et al., 2022a), because this emotion provides the strength and motivation to combat the fear and respond to the perception of injustice (Falla et al., 2025). Sympathy reflects concern for others’ misfortune (Malbois, 2023). Although closely related to empathy, these constructs differ in important ways. Whereas empathy involves sharing another person’s emotional state, sympathy is characterized by concern without directly experiencing that state. Empathy also tends to be biased toward in-group members and specific individuals, potentially hindering impartial moral judgment by narrowing focus to emotionally salient targets (Decety, 2021; Ferguson et al., 2020). In this sense, sympathy may provide a more stable foundation for moral action in contexts where empathy is limited, as it can motivate helping behavior by fostering a sense of responsibility aligned with justice and concern for others (Daniel et al., 2014; Malti et al., 2016). For this part, schadenfreude represents a distinct moral emotion that may shape bystanders’ responses in morally relevant situations. Defined as the pleasure derived from witnessing another person’s adversity, schadenfreude reflects a reaction that, although elicited by the situation of others, is not driven by concern for their well-being but rather by the indirect enjoyment of other people’s suffering (Cabrera-Vázquez et al., 2025; Van Dijk & Ouwerkerk, 2014). Schadenfreude is associated with emotional distancing and reduced prosocial motivation toward the person experiencing adversity.

Moral courage may be conceptualized in terms of its functional role in social contexts involving moral transgressions—such as bullying, harassment, or discrimination—where bystanders must choose between passivity and intervention despite potential social or personal costs. During adolescence, bullying represents one of the most frequent immoral interaction dynamics (Romera et al., 2019a). While most adolescents express moral disapproval of bullying and state that they would try to stop it and/or support the victims (Pouwels et al., 2017), fewer than half of those who witness bullying actually engage in defending acts (Ma et al., 2019). Previous qualitative studies have highlighted that witnesses in bullying situations perceive the costs associated with defending, emphasizing both social (e.g. social disapproval or exclusion) and personal costs (e.g. fear of becoming the next victim or low perceived competence) among the most frequent reasons for choosing not to intervene (Forsberg et al., 2018; Thornberg et al., 2012). Evidence further indicates that adolescents consciously evaluate these potential costs before deciding whether to intervene (Spadafora et al., 2018). From a social-cognitive perspective (Bandura, 2016), this decision can be understood as an evaluative process in which bystanders first interpret the situation as a moral transgression and then assess whether intervening is both morally required and personally feasible. During this appraisal, adolescents may engage in a form of socio-cognitive cost-benefit analysis in which anticipated risks, evaluation of their competence to intervene, and the absence of direct personal benefit are considered before action is taken (Li et al., 2025). This socio-cognitive evaluation is consistent with models of bystander decision-making in bullying, which suggest that witnesses must interpret the situation as requiring intervention and assess both personal efficacy and potential consequences before deciding to act (Nickerson et al., 2014; Pozzoli & Gini, 2010). From this perspective, acting courageously requires not only recognizing moral wrongdoing but also resisting social pressures and cognitive mechanisms that justify inaction (Thornberg et al., 2023). Moral courage is therefore more likely to emerge when the perceived moral obligation to defend the victim outweighs anticipated personal and social costs.

In this regard, moral courage does not result from a cognitive bias that underestimates risk, but from the predominance of justice and moral responsibility in this evaluative process, enabling action despite potential negative consequences. When the perceived moral imperative to defend the victim outweighs anticipated personal and social costs, adolescents may translate moral judgment into courageous action (Halmburger et al., 2016). This strong moral conviction encompasses the ability to resist social pressure, preserve self-efficacy, and persist resiliently in the face of potential consequences. Consequently, the level of activation of moral courage may predict the type of behavior adolescents adopt when they are witnesses to school bullying.

Instruments for Measuring Moral Courage

Research on the measurement of moral courage has been conducted through a variety of strategies, including experimental designs (Sasse et al., 2022b), personal previous situations (Santilli et al., 2021), moral dilemmas (Dal Cason et al., 2020), self-reports (Numminen et al., 2019), and other rating reports (Kutlaca et al., 2020).

To understand the diversity of instruments, it is important to first consider those scales in which moral courage was not directly measured. Some studies have explored moral courage as one dimension of general courage (Gibbs et al., 1986; Santilli et al., 2021). A representative example of scale is the Woodward-Pury Courage Scale (Woodward & Pury, 2007), which measures general courage and includes social–moral courage as one of its four dimensions. Other contributions come from scales that conceptualize moral courage as a kind of prosocial behavior, such as the Online Prosocial Scale (Kinnunen et al., 2016), which comprises two subscales: help-giving and moral courage. Moral courage has also been measured through instruments specifically designed for a specific context, especially in workplace setting where moral decision-making is central. For example, the Nurse Moral Courage Questionnaire (Numminen et al., 2019), which includes sub-dimensions like commitment to good patient care or genuine presence with the patient, or scales designed for digital contexts, where anonymity plays a differential role (Kinnunen et al., 2016; Pan et al., 2024). A domain-specific approach is also relevant for the study of bullying, given the unique social and personal risks faced by bystanders (Forsberg et al., 2018). While some studies have explored the moral courage of bystanders in bullying situations (Dal Cason et al., 2020; Steward, 2020), none have used instruments specifically designed to measure moral courage in this context. Using domain-specific instruments makes it possible to capture the evaluative process bystanders engage in when deciding whether to intervene, including both social costs (fear of group rejection or social exclusion) and personal costs (e.g. fear of retaliation, becoming the next victim, or feeling unprepared) associated with defending behavior (Forsberg et al., 2018; Thornberg et al., 2012). Instruments developed for this context should include items that reflect these perceived risks, as they represent the main barriers adolescents face when deciding to intervene. Considering these dimensions allows for a more accurate assessment of adolescents’ willingness to defend victims despite potential social and personal costs.

In addition to the contextual focus, another limitation of existing instruments is their predominant application to late adolescence and adulthood, leaving a notable gap in research on early and middle adolescents. One of the few instruments developed for this population is the Moral Courage Scale (Bronstein et al., 2007). This 15-item measure assesses individuals’ willingness to stand up for their values and beliefs, particularly in situations involving fairness and justice. It is organized into two dimensions: moral courage and moral reticence. The scale has been used in studies with children and adolescents (Fan & Yang, 2025). Despite its contributions, the scale presents some limitations: it adopts a broad conceptualization of moral courage by including elements of personal courage (e.g. defending oneself in unfair situations), and it does not focus on specific situations such as bullying, which is essential for capturing the complex social dynamics of these situations.

To address these limitations, this study aims to examine and validate the factor structure of the Moral Courage in Bullying Scale (MCBS), an instrument specifically designed for young adolescents and the school context, with particular attention to peer aggression (Romera et al., 2024). The development of a conceptually aligned, developmentally adapted, and context-specific instrument is essential for advancing research on the role of moral courage in predicting defensive behaviors and fostering moral reasoning among students. Supporting this approach, a recent study using the MCBS (García-Carrera et al., 2026) found that moral courage was positively associated with prosocial defending behaviors (e.g. comforting the victim or using solution-focused strategies), but not with aggressive forms of defense.

Current Study

The present study had three main objectives: (a) to develop and validate a scale for measuring moral courage specifically adapted to the context of bullying among adolescents; (b) to analyze whether the factor structure of the scale is invariant across gender (boys and girls) and developmental stage (early and middle adolescence); and (c) to examine convergent, discriminant and criterion validity of the scale through its associations with emotional processes (sympathy, moral anger and schadenfreude), prosociality and bystander behaviors.

Four sets of hypotheses were formulated. It was expected that responses would reflect adolescents’ moral courage with good psychometric properties, including factorial validity and internal reliability (Hypothesis 1). An equal conception of the construct of moral courage between boys and girls was expected (Hypothesis 2a: invariance of the scale), and according to previous studies with adolescents (Fan & Yang, 2025; García-Carrera et al., 2026), similar levels of moral courage in both genders and developmental stage were expected (Hypothesis 2b). Third, to establish convergent validity of the scale, it was hypothesized that moral courage would be moderately and positively related to the three variables: moral anger (Hypothesis 3a), sympathy (Hypothesis 3b), and prosociality (Hypothesis 3c). But there would be no link between moral courage and both dimensions of schadenfreude (Hypothesis 3d). Moral anger has been identified as a motivational force for confronting injustice and supporting victims in bullying situations (Falla et al., 2025). Sympathy could directly activate the desire to protect and care for others, thereby placing it in a closer relationship with moral courage (Malbois, 2023). Additionally, prosociality requires moral courage when the action involves potential costs (Sasse et al., 2022b). By contrast, malicious joy—whether driven by justice or aversion—is directed at others’ distress, unlike the emotions underlying moral courage (Cabrera-Vázquez et al., 2025). Lastly, criterion validity was examined, exploring the longitudinal predictive associations between moral courage and different kinds of bystander behaviors. Consistent with findings that courage enables adolescents to intervene despite fear of costs (Pouwels et al., 2019). Moral courage was expected to be positively associated with defending behaviors (Hypothesis 4a) and negatively associated with reinforcing (Hypothesis 4b) and passive (Hypothesis 4c) behaviors.

Participants

Participants were recruited as part of a National Research project about the risk and protective factors involved in bullying. Data were collected in eight secondary schools from southern Spain, 25% of which were public schools. Parental consent was obtained from the 79% of the 2,057 students who were invited to take part in the study. This resulted in a final sample of 1,628 students (48.4% girls), aged 11 to 16 years (M = 13.47 years; SD = 1.02). Participants were distributed in the first 3 grades of the Spanish secondary school: 554 (34%) students in first grade, 573 (35.2%) in second grade, and 501 (30.8%) in third grade. Socio-economic data showed that 36% of the participants came from low-income areas, 25.6% from medium-income areas, and 38.4% from high-income areas.

To cover the study’s aims, the sample was randomly divided into two equally distributed samples, based on gender and school grade. Sample 1 (n = 811 participants; 48% girls; M_age = 13.47, SD = 1.02) was used to analyze the MSBS’ factor structure, psychometric properties, internal reliability and invariance. While Sample 2 (n = 817 participants; 49% girls; M_age = 12.56, SD = 1.58) was used for confirmatory analyses and concurrent validity.

Of the total participants, 1,510 were retained for the second measurement point. Specifically, 508 students progressed from first to second grade of secondary school (5% repeated the grade), 494 progressed from second to third grade (6% repeated), and 419 progressed from third to fourth grade (7% repeated).

Measures

Procedure

Data collection was performed in May of the two academic years 2021 to 2022 and 2022 to 2023. Schools were selected based on accessibility. Parent written consent was required to participate in the survey. Participation was anonymous, confidential, and voluntary. The questionnaires were administered during regular lessons under the supervision of research staff. Data collection took about 40 min. The study design was conducted in accordance with the Declaration of Helsinki and was reviewed and approved by the Ethics and Biosafety Committee of the Universidad de Cordoba [Code: CEIH-21-17].

Item Generation of MCBS

The development of the MCBS follows the international guidelines for test construction and validation established in Standards for Educational and Psychological Testing (AERA et al., 2014), and the methodological steps proposed for the design of evaluation instruments (Muñiz & Fonseca-Pedrero, 2019). The process was based on the framework proposed by Boateng et al. (2018). The format of the items and the response options were defined according to psychometric writing recommendations to ensure clarity and validity (Moreno et al., 2015). First, a comprehensive review of scientific literature was conducted to establish a well-founded definition of moral courage in bullying situations. Moral courage was conceptualized as the deliberate decision by a witness to act according to moral values despite personal or social costs. Item development was guided by motivational themes identified in previous qualitative studies (Forsberg et al., 2018; Thornberg et al., 2012), which showed that adolescents consider both personal risks (e.g. fear of becoming a victim, lack of confidence) and social risks (e.g. opposing group norms, social rejection) when deciding whether to intervene. By incorporating these dimensions, the instrument aims to capture adolescents’ willingness to act in defense of victims despite potential risks, providing a realistic and context-sensitive assessment of moral courage in bullying situations. As a result, an initial set of 10 items was developed. All items were phrased in terms of behavioral willingness (e.g. “I would defend a bullying victim even if. . .”) rather than perceived capability (“I could. . .”), to capture adolescents’ readiness to act according to moral values despite anticipated personal or social costs.

Next, the content validity was assessed with the target population. Four focus groups with 40 secondary school students (57.5% girls; M_age = 13.48, SD = 1.24) were conducted in 2 schools located in Cordoba (Spain), which did not participate in the study. Informed consent from participants and their parents was obtained. Sessions were audio-recorded, transcribed, and analyzed. Two trained researchers guided the focus groups to ensure participants understood the items. This process led to linguistic, semantic, and functional revisions, refining and optimizing the original items. As a result, some words and the structure of the items were modified. Subsequently, three expert judges reviewed the revised items to reduce bias and ensure clarity, parsimony, and internal consistency. As a result, two items were removed due to conceptual imprecision (items 2 and 6 in Table S1). Item 2 was removed because expert reviewers found that it emphasized potential reputational benefits rather than the moral willingness to act despite social costs. This focus could introduce ambiguity regarding the moral motivational basis of behavior. Item 6 was removed because experts consider it complex and unclear regarding the specific cost it referred to. A pilot study was then conducted with 202 secondary students (50% girls; M_age = 13.24, SD = 0.98) to examine the psychometric properties of the remaining 8 items using exploratory factor analysis (EFA). Two additional items with low factor loadings (<0.40) were eliminated (items 5 and 10 in Table S1). As a result of this refinement process, the final version of the MCBS retained six items, three related to personal costs (retaliation, perceived efficacy, and personal gain) and three about social costs (group disapproval, low relational obligation with the victim, and social rejection or exclusion) in bullying defending (see Table S1 to find the specific retained item associated to each kind of cost). Lastly, the scale was applied to the study sample.

Data Analysis

The factor structure of the MCBS was explored through EFA in the Sample 1 and confirmatory factor analysis (CFA) in the Sample 2. The Kaiser–Meyer–Olkin (KMO) sampling adequacy measure (>0.6 value) and Bartlett’s test of sphericity (Watson, 2017) were run to determine the appropriateness of applying the EFA (Kaiser, 1974). These analyses were conducted in SPSS 28.0. The number of factors were established when the eigenvalues exceeded one (Kaiser–Guttman criterion) and were greater than the values extracted from Horn’s parallel analysis (Franklin et al., 1995). Items with communalities and factor loadings lower than 0.40 were excluded (Costello & Osborne, 2005).

A CFA was conducted to confirm the goodness of fit of the factor model proposed by the AFE. Model fit was evaluated using Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) —≥0.90 for acceptable and ≥0.95 for good—and the Root mean square error of approximation (RMSEA) and the Standardized Root Mean Square Residual (SRMR)—≤0.08 for acceptable and ≤0.06 for good—(Chen, 2007). Internal reliability of factors was estimated through Cronbach’s α, McDonald’s ω, and composite reliability (ρc), with >.7 considered acceptable (Geldhof et al., 2014). Average variance extracted (AVE) was calculated to check the external discriminant validity (≥0.5; Hair et al., 2010). Test-retest reliability (1 year later) was calculated using the intraclass correlation coefficient (ICC _(A,1)), reflecting the consistency of individual scores across repeated measurements (Shrout & Fleiss, 1979).

Measurement invariance was analyzed across gender (boys vs. girls), developmental stage (early vs. middle adolescence), and time (T1 vs. T2) to identify whether the factor structure was equivalent across genders, developmental stage, and whether it measured the same construct over time. For this purpose, a series of increasingly restrictive nested models (configural vs. metric and metric vs. scalar) were performed, and it was verified that no more than two fit criteria were violated: ΔCFI < .01, ΔRMSEA < .015, and Δχ² > .05, which indicated that the more restrictive models maintain good metric and scalar equivalence between groups. Invariance and latent means were compared between Samples 1 and 2 to examine the lack of bias in the measurement of the construct.

Data from both samples were combined to explore concurrent and longitudinal validity. Gender differences were analyzed using t-tests, with effect sizes estimated by Cohen’s d (Cohen, 1992). Concurrent and discriminant validity were examined using Pearson correlations with conceptually related constructs and with conceptually unrelated constructs. Longitudinal validity was tested using hierarchical linear regression analyses to determine whether moral courage at Time 1 predicted the kind of bystander behaviors adopted in a bullying situation. In the first step, gender, age, and bystander behavior on Time 1 were included as control variables. In the second step, moral courage in Time 1 was included as a predictor of bullying bystander behaviors (Time 2). Regression assumptions were checked. Multicollinearity was tested using the variance inflation factor (VIF; 1–10 indicates no multicollinearity); and tolerance measures (T > 0.40 as accepted). The Durbin-Watson statistic was run to check correlation errors; values from 1.5 to 2.5 indicate independent errors (Durbin & Watson, 1971).

EFA, CFA, and invariance analyses were developed in Mplus Version 8.11 (Muthén & Muthén, 1998) with maximum likelihood robust estimator as recommended for five responses categories (Rhemtulla et al., 2012). All validity analyses were performed in SPSS 28.0. (IBM Corp., 2021) Statistical significance was set at p < .05. This research was not preregistered. Data, analytic code, and materials associated with this manuscript are not publicly accessible.

Exploratory Factorial Analysis

In Sample 1, the KMO value (0.87) and the statistical significance of Bartlett’s test of sphericity, χ² (15) = 1,583.661, p < .001, supported that EFA was appropriate for the screening of the MCBS factor structure. The eigenvalue criteria indicated that a one-factor structure was the most appropriate (3.46). Horn’s parallel analysis confirmed the structure because the eigenvalue in EFA for two factors (0.72) was lower than the eigenvalue in parallel analysis (1.07). The single-factor model explained the 57.7% of the total variance. All items had factor loading greater than 0.40 (see Table S2 in the Supplemental Material).

Descriptive statistics showed that the first item (“I would stand up in a bullying situation, even if most people disapprove”) had the highest mean scores (M = 3.8), while the lowest levels (M = 3.25) were reported on item 4 (“I would defend a bullying victim, even if my classmates turn their backs on me for it”). Only MCBS4 and MCBS6 showed slight deviation in skewness (K = −1.13 and K = −1.08, respectively). The bivariate correlations revealed a moderate and significant association among all items (from r = .4 to .63), which confirmed the absence of multicollinearity among them (see Table S3 in the Supplemental Material).

CFA and Measurement Invariance Analyses

The CFA of the unidimensional structure of the scale showed optimal fit indices: χ² (9) = 39.196, p < .001, CFI = 0.972, TLI = 0.953, RMSEA = 0.068, 90% CI [Confidence interval] [0.047, 0.091], SRMR = 0.028. Factor loadings were statistically significant, greater than 0.70, and low in measurement errors (see Figure 1). This factor had good internal reliability: Cronbach’s α = .87 (CI [0.85, 0.88]); McDonald’s ω = .87 (CI [0.85, 0.88]); ρc = .87. The AVE was 0.52, indicating acceptable convergent validity for a single-factor model. The six retained items addressed 43.25% of the variance of the model. The ICC indicated a good test-retest reliability (ICC_(A,1) = 0.71, 95% CI [0.67, 0.74]). This result confirmed an adequate stability and reproducibility of the instrument scores in two separate measurement moments.

OPEN IN VIEWER

Figure 1.

Confirmatory factor analysis for the one-dimensional MCBS model.

The results of the multi-group analysis showed that the unrestricted configural invariance model had a good baseline fit for gender, developmental stage, and time. Measurement invariance results (ΔCFI/ΔTL < .01 and ΔRMSEA < .015) showed that the one-factor-structure of the 6-items MCBS was invariant across gender, developmental stage, and longitudinally (see Table 1).

OPEN IN VIEWER

Table 1.

Multi-Group Invariance of MCBS across Gender.

Model tested	Model fit indices				Model comparison
	χ2 (df)	CFI	TLI	RMSEA [90% CI]	ΔCFI	ΔTLI	ΔRMSEA
Gender: boys vs. girls
Configural	45.364 (18)*	0.974	0.954	0.065 [0.042, 0.089]	—	—	—
Metric	54.171 (23)*	0.970	0.961	0.061 [0.040, 0.083]	.004	.007	.004
Scalar	66.995 (28)*	0.963	0.960	0.062 [0.043, 0.082]	.007	.001	.001
Age: early vs. middle
Configural	49.318 (18)*	0.971	0.951	0.07 [0.047, 0.093]	—	—	—
Metric	58.183 (23)*	0.967	0.957	0.065 [0.045, 0.086]	.004	.006	.005
Scalar	64.05 (28)*	0.966	0.964	0.06 [0.041, 0.079]	.001	.007	.005
Time: T1 vs. T2
Configural	75.48 (18)*	0.974	0.957	0.068 [0.053, 0.085]	—	—	—
Metric	85.929 (23)*	0.972	0.963	0.063 [0.049, 0.078]	.002	.004	.005
Scalar	103.867 (28)*	0.966	0.963	0.063 [0.050, 0.076]	.006	.000	.000

Note. MCBS = Moral Courage in Bullying Scale; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA = Root mean square error of approximation.

*

p < .001.

Concurrent, Discriminant, and Longitudinal Validity

Students’ t-test showed that girls had significantly higher levels of moral courage than boys, as well as of moral anger, sympathy, and prosociality. While boys had significantly higher levels of schadenfreude for aversion. All these results were replicated on Time 2 (see Table 2). Students’ t-test was not significant for comparative analyses between early and middle adolescents for all variables, except schadenfreude for aversion (t = 4.531, df = 1,121, p < .001, d = .39). Specifically, early adolescents had higher levels than middle adolescents (T1: M = 1.84, SD = 0.94 vs. M = 1.6, SD = 0.79; T2: M = 1.82, SD = 0.99 vs. M = 1.58, SD = 0.85). Correlation analyses revealed that moral courage was positively associated with all analyzed variables (moral anger: r = .44, p < .001; sympathy: r = .4, p < .001; prosocial behaviors: r = 0.39, p < .001), which suggests that these constructs are related but not equivalent. Both dimensions of schadenfreude were not correlated to moral courage (see Table 2).

OPEN IN VIEWER

Table 2.

Descriptive, Gender Comparative and Correlations Analysis for Concurrent and Discriminant Validity.

Variables	Descriptive (SD)			Student’s t-test		Correlations
	Total sample	Boys	Girls	t(gl)	d	1	2	3	4	5	6	7	8	9	10	11
1. MC T1	3.54 (0.97)	3.42 (1.01)	8.67 (0.89)	−4.748 (1,332.6)***	−.26
2. MA T1	3.65 (1.13)	3.41 (1.18)	3.92(0.99)	−8.394 (1,252.4)***	−.47	.42***
3. S T1	4.1 (0.81)	3.91 (0.87)	4.1 (0.68)	−9.418 (1,262.7)***	−.52	.39***	.59***
4. PB T1	4.01 (0.97)	3.87(0.99)	4.17 (0.93)	−5.563 (1,298)***	−.31	.37***	.52***	.52***
5. SA T1	1.58 (0.81)	1.65 (0.86)	1.51 (0.74)	3.165 (1,264.3)***	.18	.00	−.01	−.06*	−.08**
6. SJ T1	3.56 (1.11)	3.55 (1.12)	3.57 (1.09)	−0.273 (1,286)	−.02	−.01	.08**	.04	.02	.06*
7. MC T2	3.43 (0.98)	3.28 (1.01)	3.6 (0.91)	−5.728 (1,172.6)***	−.33	.56***	.4***	.37***	.37***	−.03	−.04
8. MA T2	3.44 (1.18)	3.16 (1.18)	3.75 (1.1)	−8.878 (1,202.7)***	−.51	.26***	.44***	.35***	.28***	−.05	.04	.36***
9. S T2	4.02 (0.85)	3.81 (0.9)	4.25 (0.69)	−9.774 (1,203.9)***	−.55	.31***	.44***	.46***	.36***	−.05	.00	.42***	.56***
10. PB T2	3.89 (1.04)	3.75 (1.08)	4.06 (0.98)	−5.283 (1,235.5)***	−.3	.28***	.35***	.31***	.4***	−.05	.00	.37***	.48***	.55***
11. SA T2	1.62 (0.88)	1.68 (0.92)	1.55 (0.82)	2.6633 (1,238.1)**	.15	−.04	−.04	−.04	−.04	.32**	.05	−.05	−.09**	−.08*	−.08*
12. SJ T2	3.39 (1.18)	3.38 (1.19)	3.39 (1.17)	−0.155 (1,232)	−.01	−.03	.08*	.05	.04	.01	.37***	−.02	.03	.03	.02	.15***

Note. MC = Moral Courage; MA = Moral Anger; S = Sympathy; PB = Prosocial Behavior; SA = Schadenfreude for Aversion; SJ = Schadenfreude for Justice; SD = Standard deviation.

*

p < .05. **p < .01. ***p < .001.

Preliminary analyses showed that there was no evidence of multicollinearity for any of the bystander behaviors: bully reinforcer (T = from 0.96 to 0.99, VIF = from 1.01 to 1.04), victim defender (T = from 0.59 to 0.99, VIF = from 1.01 to 1.7) and outsider (T = from 0.77 to 0.99, VIF = from 1.01 to 1.31). The Durbin-Watson statistics were 1.95 for bully reinforcing, 1.93 for victim defending, and 1.88 for outsider behaviors, indicating that the errors were independent among the independent variables. The results of the three hierarchical linear regression models, one for each kind of bystander behavior, showed that moral courage did not have a significant effect on reinforcing behaviors. But it had a positive effect on defending behaviors, while a negative effect was found for outsider behaviors (see Table 3).

OPEN IN VIEWER

Table 3.

Hierarchical Regression Analyses Predicting Three Kinds of Bullying Bystander Behaviors.

Predictor variables	Bully reinforcer				Victim defender				Outsider
	β	SE	βe	T	β	SE	βe	t	β	SE	βe	t
Step 1
Gender	−0.27	0.08	−0.17	1.04	0.43	0.12	0.14	3.43***	−0.06*	0.12	−0.02	−0.53
Age	0.03	0.03	0.05	−3.5***	−0.04	0.04	−0.04	−0.93	−0.05**	0.04	−0.01	−0.11
BB	0.33	0.06	0.28	5.89***	0.61	0.05	0.53	12.93***	0.53	0.05	0.5	11.64***
R2	0.11				0.32				0.24
Step 2
MC	−0.03	0.04	−0.04	−0.83	0.22	0.08	0.13	2.69**	−0.18	0.05	−0.18	−3.8***
∆R2	.002				.01***				.025***
R2	.111				.33				.26

Note. 1 = Boys; 2 = Girls. BB = Bystander Behavior; MC = Moral Courage; SE = Standard error.

*

p < .05. **p < .01. ***p < .001.

Limitations and Practical Implications

These findings must be considered in light of several limitations. First, the use of non-probabilistic sampling restricted the study to adolescents from a single region in Spain. Moreover, the sample was large and homogeneous in age and gender, but no data were collected on nationality or other cultural indicators, so it is not possible to determine whether perceptions and expressions of moral courage would be consistent in different cultural contexts. Cross-cultural studies are needed to test the scale’s generalizability across diverse demographic and cultural contexts. In addition, the influence of other factors such as race, religion, socioeconomic status, or disabilities should be explored on the scale’s reliability and invariance. Second, the study relied on self-reports, which capture personal moral-cognitive processes and are less affected by reputational biases; but may be affected by social desirability bias and participants’ mood. Combining self-reports with peer or teacher reports could improve measurement accuracy. Thirdly, criterion validity with other existing moral courage scales was not assessed; future research should explore its convergence with other instruments. Lastly, although the present study examines criterion validity of the scale through its associations with emotional processes (sympathy and moral anger), the construct may partially overlap with other related concepts that also involve helping under potential social costs, such as prosocial risk-taking (Armstrong-Carter & Telzer, 2025). While moral courage emphasizes a moral–normative motivation to confront wrongdoing, prosocial risk-taking describes different ways to provide meaningful, practical, and emotional support to others, even when doing so involves personal risk. A relevant difference between both is that moral courage places greater emphasis on the moral-normative motivation underlying the behavior, rather than solely on the risk structure of the decision. Future research should examine and clarify these conceptual boundaries more directly by incorporating comparative measures of prosocial risk-taking and testing discriminant validity through multi-construct designs.

The development of a robust and psychometrically valid scale to assess moral courage in bullying situations represents a significant advancement for both research and educational practice. Having a reliable measure of moral courage facilitates the detection of student profiles more likely to intervene as defenders, thereby helping to disrupt power dynamics that perpetuate bullying. The application of this scale in schools could provide key information for the design of educational public policies and school-based programs, strengthening the development of socio-emotional and moral competence, critical thinking, and an ethics of care. Understanding adolescents’ moral courage in witnessing injustice is crucial for fostering ethical, civic, and prosocial behavior, contributing to a more sustainable and peaceful society.

Conclusions

This study provides new evidence by introducing a specific scale with strong psychometric properties for assessing moral courage in school bullying contexts. This tool enables a more contextualized and realistic exploration of its association with bystander behaviors. Moreover, the scale demonstrates both gender and longitudinal invariance, supporting its reliability and validity, and captures the distinct nature of moral courage in relation to other constructs such as moral anger, sympathy, and prosociality. These features enable more nuanced research into the complexity of moral-cognitive processes involved in bullying. Longitudinal analyses showed that moral courage predicts increases in defending behaviors and decreases in passive bystander behaviors over time, highlighting its role in shaping adolescents’ proactive engagement in bullying situations. In contrast, the lack of association with reinforcing behavior suggests the need to explore additional moral, motivational, and cognitive mechanisms to fully understand the multifaceted nature of bullying dynamics.