The sample consisted of employed women (i.e., individuals who self-identified as women) from a wide range of occupational sectors and Spanish regions. A non-probabilistic sampling strategy was employed, aimed at maximizing occupational and territorial heterogeneity, as commonly recommended in psychometric validation studies to examine factorial stability and external validity in adapted instruments (Boateng et al., 2018; Worthington and Whittaker, 2006).

The inclusion criteria were: (a) self-identifying as a woman, (b) being 18 years of age or older, and (c) being actively working at the time of data collection. A total of 788 participants initially met the inclusion criteria. Cases with incomplete responses on the MIMI-16 were excluded using listwise deletion, resulting in the exclusion of 32 cases (Sample 1: n = 10; Sample 2: n = 22) and a final analytic sample of N₁ = 352 and N₂ = 404 participants. The proportion of excluded cases was low (approximately 4.1%), suggesting a limited impact on the representativeness of the analytic sample.

Data collection was conducted using an online self-administered questionnaire hosted on a secure digital platform. The study invitation was disseminated through collaborating organizations, professional networks, and social media channels. Participation was voluntary and anonymous, and all participants provided informed consent prior to completing the survey. Data were collected at two different time points for each sample, and recruitment was conducted through partially independent dissemination channels. No participant overlap between samples was allowed or detected.

We assessed (a) the target measure, the original MIMI-16 scale, and (b) a set of external criterion measures used to examine criterion-related and nomological validity evidence.

To examine criterion-related and nomological validity evidence, the MIMI-16 was correlated with a set of psychological and organizational variables that have been theoretically and empirically linked to experiences of gender microaggressions in the workplace.

Preliminary data screening and assumption checking—including the examination of missing values, univariate normality, skewness, and kurtosis—were conducted using IBM SPSS Statistics (Version 27). Missing data were handled using listwise deletion, because only participants who completed all MIMI-16 items were retained for analysis. Therefore, CFA and multigroup CFA models were estimated using complete cases, and no imputation procedures were applied. CFA were estimated using the robust weighted least squares estimator with mean and variance adjustment (WLSMV), which is recommended for ordinal indicators. Given the six-point rating-type response format of the MIMI-16 and the observed deviations from normality in several items, WLSMV provides more appropriate parameter estimates and standard errors than maximum likelihood estimators. Although the original study employed MLR, the present study adopted WLSMV to better accommodate the ordinal nature of the data.

The model was first estimated using the combined sample (N = 756) and subsequently examined in two independent subsamples (N = 352 and N = 404) to assess the stability of the factorial structure across samples. In addition to the hypothesized correlated two-factor model, alternative theoretically plausible models (i.e., unidimensional, higher-order, and bifactor models) were also estimated and compared to evaluate the adequacy of the factorial structure. Model fit was evaluated using multiple indices: the chi-square statistic (χ²), the Comparative Fit Index (CFI), the Tucker–Lewis Index (TLI), the Root Mean Square Error of Approximation (RMSEA) with 90% confidence intervals, and the Standardized Root Mean Square Residual (SRMR), following commonly accepted criteria (Hu and Bentler, 1999).

Given that the MIMI-16 has a clearly specified and empirically supported factorial structure in its original version (Algner and Lorenz, 2022), the present study adopted a confirmatory approach. Accordingly, no exploratory factor analysis was conducted, as the aim was not to identify an unknown latent structure but to evaluate the replicability and stability of a previously specified theoretical model in the Spanish workplace context. This decision is consistent with current methodological recommendations, which discourage the use of exploratory analyses when strong theoretical and empirical foundations exist and instead prioritize confirmatory methods for testing validated models (Boateng et al., 2018; Flora and Flake, 2017; Worthington and Whittaker, 2006).

Standardized factor loadings were computed, and internal consistency was assessed using Cronbach’s alpha (α), McDonald’s omega (ω), and composite reliability (CR). Evidence of convergent and discriminant validity was examined through the average variance extracted (AVE) and the Fornell–Larcker criterion (Fornell and Larcker, 1981). Additionally, the heterotrait–monotrait ratio (HTMT) was computed as a complementary indicator of discriminant validity. These analyses were conducted using the semTools package (Jorgensen et al., 2021) and the MBESS package in R.

Finally, the results of the original validation study by Algner and Lorenz (2022) were used as a comparative benchmark for interpreting model fit and psychometric performance.

In addition to Pearson bivariate correlations used to examine criterion-related validity at the scale level, supplementary Spearman rank-order correlations were computed to assess the robustness of the observed associations given the ordinal nature of the response formats. This sensitivity analysis allowed us to verify that the pattern of relationships did not depend on parametric assumptions.

Item-level descriptive statistics for the MIMI-16 are presented in Table 4. Across both samples, most items showed pronounced positive skewness and elevated kurtosis, indicating a strong concentration of responses at the lower end of the scale. Several items, particularly those reflecting more explicit forms of microinsults, exhibited substantial departures from normality, consistent with the presence of floor effects.

Following the methodological approach adopted in the original validation study of the MIMI-16 (Algner and Lorenz, 2022), Pearson bivariate correlations were first computed to examine the pattern of associations between MIMI-16 scores and the criterion variables related to psychosocial functioning and organizational outcomes.

Although the MIMI-16 items are ordinal, correlations were computed on aggregated scale scores, which approximate continuous distributions. To ensure robustness of the findings, supplementary analyses using Spearman rank-order correlations were also conducted. These yielded an equivalent pattern of associations in terms of direction, magnitude, and statistical significance across both samples (see Supplementary Tables S1, S2), supporting the stability of the results.

As an external organizational indicator, a global index of social relations and support was included. This index was derived from the FPSICO 4.1 psychosocial risk assessment instrument and computed by aggregating items assessing social relations, support, and exposure to problematic interpersonal interactions at work (see notes to Tables 5, 6). Additional correlations were examined with indicators of psychological health and work-related psychosocial variables. For Sample 1, these included psychological distress, work engagement, job satisfaction, and burnout (emotional exhaustion, depersonalization, and personal accomplishment). For Sample 2, analyses focused on core self-evaluations, turnover intention, and general self-efficacy.

Following the original two-factor structure proposed by Algner and Lorenz (2022), a correlated two-factor model was specified, comprising microinvalidations (mi_1–mi_8) and microinsults (mi_9–mi_16). Given the ordinal nature of the response scale and the marked skewness observed in several items, models were estimated using WLSMV.

The model was first tested separately in each sample and subsequently in the pooled dataset. Fit indices for the correlated two-factor model across samples are presented in Table 7. In Sample 1 (N = 352), the model showed overall acceptable fit, although RMSEA values were slightly above conventional thresholds. In Sample 2 (N = 404), fit indices were good or very good. In the combined sample (N = 756), the model demonstrated excellent overall fit, χ² (103) = 223.09, p < 0.001, CFI = 0.983, TLI = 0.981, SRMR = 0.057, and RMSEA = 0.051, 90% CI [0.045, 0.057].

To further evaluate the dimensional structure of the instrument, several theoretically plausible alternative models (i.e., unidimensional, higher-order, and bifactor models) were estimated and compared. The corresponding fit indices are presented in Table 8. Comparative analyses indicated that the unidimensional model provided inferior fit, whereas the higher-order model did not yield interpretable parameter estimates. Although the bifactor model showed improved nominal fit, it resulted in inadmissible solutions (negative latent variances), suggesting model instability and potential overfitting.

Taken together, these results support the correlated two-factor model as the most parsimonious and interpretable solution, while acknowledging the substantial shared variance between factors.

To examine the stability of the MIMI-16 measurement model across independent samples of working women, multigroup CFA were conducted using the WLSMV estimator, which is recommended for ordinal indicators. The results are presented in Table 10.

Measurement invariance was tested sequentially following standard procedures for ordinal data: configural invariance (same factorial structure across groups), metric invariance (equal factor loadings), and scalar invariance (equal factor loadings and thresholds). Because WLSMV estimation does not rely on traditional χ² difference testing, decisions were based on changes in approximate fit indices, in line with current methodological recommendations (Cheung and Rensvold, 2002; Chen, 2007). Specifically, invariance was supported when changes in CFI were ≤0.010, RMSEA ≤0.015, and SRMR ≤0.030 for metric and ≤0.010 for scalar invariance.

The configural model demonstrated acceptable fit (CFI = 0.984; RMSEA = 0.048; SRMR = 0.065), indicating that the two-factor structure was similarly represented in both samples. Imposing equality constraints on factor loadings (metric invariance) did not result in a meaningful deterioration of model fit (CFI = 0.987; RMSEA = 0.043; SRMR = 0.075), with changes in fit indices (ΔCFI = 0.003; ΔRMSEA = −0.005; ΔSRMR = 0.010) well below recommended thresholds. Likewise, the scalar invariance model (equal loadings and thresholds) showed comparable fit (CFI = 0.984; RMSEA = 0.046; SRMR = 0.065), again meeting established criteria for invariance.

Taken together, these findings support configural, metric, and scalar invariance of the MIMI-16 across samples, indicating that the factorial structure, factor loadings, and item thresholds are largely comparable. This level of invariance supports comparisons of structural relations across samples. In the present study, scalar invariance was primarily established to confirm measurement stability rather than to support latent mean comparisons.

Internal consistency was high for both dimensions. Ordinal alpha values were 0.92 (microinvalidations) and 0.95 (microinsults). Model-based omega coefficients were 0.88 and 0.89, respectively. Composite reliability (CR) values exceeded 0.93 for both factors, indicating strong internal consistency.

Convergent validity at the latent level was examined through the average variance extracted (AVE). AVE values were 0.64 for microinvalidations and 0.73 for microinsults, exceeding the recommended 0.50 threshold and suggesting adequate to strong convergence between indicators and their respective latent constructs.

The latent correlation between factors was high (r ≈ 0.83), reflecting the conceptual proximity of both dimensions. Discriminant validity was evaluated using the Fornell–Larcker criterion. Specifically, the AVE for microinvalidations (0.64) was lower than the squared inter-factor correlation (r² ≈ 0.69), indicating that the criterion was not met for this dimension. To complement this analysis, the heterotrait–monotrait ratio (HTMT) was computed, yielding a value of 0.83, which falls below the recommended threshold of 0.85 and provides additional support for discriminant validity. The results are presented in Table 11.

Taken together, these results indicate substantial shared variance between microinvalidations and microinsults. However, considering the HTMT results and the theoretical relatedness of both dimensions within the broader microaggression framework, the distinction between factors remains interpretable, albeit with caution.

Beyond latent-level evidence, the pattern of associations with external psychosocial and organizational variables provides criterion-related and nomological validity evidence, rather than convergent or discriminant validity in the strict psychometric sense.

To contextualize the findings, descriptive comparisons were conducted between the Spanish and German validation studies. Across datasets, global fit indices were broadly comparable, with Spanish samples showing CFI and RMSEA values within—and in some cases exceeding—those reported in the original German validation. The results are presented in Table 12.

These comparisons should be interpreted with caution, as they are descriptive in nature and do not reflect formal statistical equivalence testing across samples or studies.

Nevertheless, the overall similarity in model fit provides preliminary support for the structural robustness of the MIMI-16 across different linguistic and cultural contexts.

Model fit across both independent samples and the pooled dataset supported the adequacy of the correlated two-factor structure under WLSMV estimation. Standardized factor loadings were generally moderate to high, indicating that items functioned consistently as indicators of their respective dimensions. Microinvalidations showed slightly greater variability in loadings than microinsults, which may reflect the broader and more ambiguous nature of invalidating experiences in workplace interactions. This pattern is consistent with theoretical accounts describing microinvalidations as subtle, often normalized practices that may be differentially interpreted depending on organizational culture and role expectations (Sue, 2010; Nadal and Haynes, 2012).

With respect to reliability, both dimensions demonstrated high internal consistency, as indicated by Cronbach’s alpha, McDonald’s omega, and ordinal alpha coefficients. Composite reliability and average variance extracted values further supported the convergent validity of the scale. However, the AVE for microinvalidations was lower than the squared inter-factor correlation, indicating that the Fornell–Larcker criterion was not met for this dimension. This result reflects the substantial empirical overlap between microinvalidations and microinsults, which was further qualified by HTMT values below recommended thresholds, supporting a cautious but defensible distinction between factors. In line with these findings, alternative factorial representations (unidimensional, higher-order, and bifactor models) were examined but either showed inferior fit or yielded inadmissible or unstable solutions.

Taken together, these results suggest that the two dimensions capture differentiated yet highly interrelated manifestations of gender microaggressions. This pattern is consistent with theoretical perspectives that conceptualize microaggressions as a unified phenomenon expressed through distinct but overlapping forms (Sue, 2010), thereby supporting the retention of a correlated two-factor structure despite the high degree of shared variance.

Importantly, multigroup analyses supported configural, metric, and scalar invariance across samples, indicating that the factorial structure, loadings, and thresholds operated comparably in both subsamples. This strengthens confidence in the structural stability of the instrument and supports the comparability of latent scores across independent samples within the same cultural context.

Beyond internal structure, the pattern of associations between MIMI-16 scores and external criteria provided further support for construct validity. Higher exposure to gender microaggressions was consistently associated with poorer workplace social relations, greater psychological distress and burnout, lower work engagement and job satisfaction, and higher turnover intention.

Importantly, the strength of these associations differed across samples, with stronger relationships observed in Sample 1, which included more proximal, state-like indicators (e.g., burnout, distress, job satisfaction), and weaker associations in Sample 2, which relied on more distal and dispositional constructs (e.g., core self-evaluations and occupational self-efficacy).

This pattern was robust to alternative (nonparametric) correlation estimates, with supplementary analyses yielding an equivalent pattern of results across both independent samples. These findings align with prior research documenting the cumulative psychosocial costs of gender microaggressions in organizational settings (Basford et al., 2014; Kim and Meister, 2023) and reinforce the scale’s nomological validity within occupational research contexts.

Taken together, the present findings provide converging evidence supporting the reliability, structural validity, measurement invariance, and criterion-related validity of the Spanish version of the MIMI-16. While some psychometric indicators suggest substantial overlap between dimensions, the overall pattern of results supports its use as a theoretically grounded measure of workplace gender microaggressions in both research and applied contexts.

Several limitations of the present study should be acknowledged. First, the use of a non-probabilistic sampling strategy limits the generalizability of the findings beyond the populations and occupational contexts represented in the study. Although efforts were made to maximize heterogeneity in terms of sectors, regions, and employment conditions, future research should replicate these findings using probabilistic or stratified sampling designs.

Second, the cross-sectional design precludes causal inferences regarding the relationships between gender microaggressions and psychosocial or occupational outcomes. Longitudinal designs would be particularly valuable for examining the temporal dynamics and cumulative effects of workplace microaggressions, as well as their potential role in the development of psychological distress, burnout, and turnover intention over time.

Third, all variables were assessed using self-report measures collected within the same survey context, which may raise concerns regarding common-method variance. Although the primary aim of the present study was the psychometric validation of a measurement instrument rather than causal modeling, and self-report methodologies are standard in scale validation research, future studies should consider incorporating multi-source assessments (e.g., supervisor ratings or objective organizational indicators) or longitudinal designs to further examine the robustness of the observed associations. Nevertheless, the replication of the expected pattern of associations across two independent samples reduces the likelihood that the findings are solely attributable to shared method variance.

Fourth, although the scale demonstrated acceptable levels of discriminant validity according to HTMT criteria, the high shared variance between microinvalidations and microinsults suggests that their distinction should be interpreted with caution. This overlap may reflect partially overlapping manifestations of gender microaggressions rather than fully distinct constructs, and future research should further examine the dimensional structure of the instrument using alternative modeling approaches and diverse samples.

Fifth, the study did not include established Spanish-language measures assessing conceptually related constructs (e.g., micromachismos or ambivalent sexism), which limited the extent to which convergent validity could be examined at the construct level. Incorporating such measures in future studies would strengthen convergent validity evidence.

Sixth, different sets of criterion variables were used across samples, including more proximal, state-like outcomes in Sample 1 and more distal, dispositional constructs in Sample 2. Although this design allowed us to examine the scale across diverse nomological contexts, it limited the direct comparability of criterion-related validity evidence between samples. Future studies should include overlapping criterion measures across samples to enhance comparability.

Finally, although the MIMI-16 was originally developed and validated in Germany, its application in different cultural contexts requires caution. When using the scale in other national settings, future research should consider that the specific manifestations of gender microaggressions may vary as a function of cultural and organizational context. Although the present study validated a Spanish-language version of the MIMI-16, data were collected exclusively in Spain; therefore, further research should adapt and examine the measure in other Spanish-speaking cultural contexts.

These findings should be interpreted within the constraints of cross-sectional self-report data and require replication across different cultural and organizational contexts.