Psychometric validation of the Short Dark Tetrad (SD4): an item-level re-examination with measurement invariance

Psychometric evaluations of the Short Dark Tetrad (SD4; [33]) have been conducted predominantly in Western samples, and the limited evidence from non-WEIRD contexts has yielded inconsistent findings regarding its factor structure. The only prior Turkish adaptation reported that seven items did not function adequately and were removed following exploratory factor analysis, leaving open whether this reflected genuine cultural unsuitability or methodological choices in the analysis of ordinal data. The present study re-examined the SD4 in a community sample of 417 Turkish adults (54.4% female, M age = 30.4, SD = 12.5) using confirmatory factor analysis with the WLSMV estimator and polychoric correlations, the approach recommended for ordinal Likert data. At the item level, only one item (m1, “It is not wise to let people know your secrets”) showed a substantive problem (standardized loading = .14, with cross-loadings on Psychopathy and Sadism). The six remaining items removed by Aytaç [1] performed adequately in the present sample (loadings = .53–.74). A 27-item correlated four-factor model excluding m1 showed acceptable fit (CFI = .913, TLI = .904, RMSEA = .069 [.064, .074], SRMR = .070) and outperformed a unidimensional model. Multi-group analyses across gender supported configural, threshold (weak), and threshold-plus-loading (strong) invariance (all ΔCFI ≤ .005). Convergent validity was demonstrated through theoretically expected associations with the NPI-13, MPS, LSRP, and Violence Tendency Scale ( r = .40–.69). Men scored higher than women on all four traits; effect sizes ranged from small for Narcissism (η 2 p = .010) to moderate–large for Sadism (η 2 p = .115, Cohen’s d = 0.72). The findings indicate that the original SD4 structure is largely retained in Turkish samples when ordinal-appropriate estimation is used, with Sadism emerging as the most strongly gender-differentiated trait.