Video - The Integrated Spiritual Intelligence Scale (ISIS): Examining Psychometric Properties and Measurement Invariance in a Sample of Romanian Employees

The 83-item Integrated Spiritual Intelligence Scale (ISIS) is one of the most widely used psychological tools for measuring spiritual intelligence in academic studies and real-world settings. However, its factor structure has been subject to some debate. To date, no study has been carried out using a sample of Romanian employees to test the reliability and validity of the ISIS or its factorial structure. Data were collected from twelve different study groups, involving a total of 342 Romanian employees. Using a one-stage sampling process, exploratory factor analysis (EFA) was conducted on a randomly selected pilot sample (N=168), followed by confirmatory factor analysis (CFA) on an independent group (N=174). The EFA identified three factors by Kaiser's criterion, accounting for 58.4% of the total variance across all items. Three competing measurement models were tested with CFA. Model 1, a unidimensional solution with all 21 items loading on a single factor, fit poorly, χ²=877.205, df=189, χ²/df=4.641, CFI=.660, TLI=.622, RMSEA=.145, SRMR=.127. Model 2, a correlated three-factor model that distinguishes Consciousness, Transcendence, and Truth, showed substantially better fit, χ²=379.748, df=186, χ²/df=2.042, CFI=.904, TLI=.892, RMSEA=.077, SRMR=.067. Model 3, a bifactor specification with one general factor plus three orthogonal specific factors, yielded the best global fit indices, χ²=300.371, df=168, χ²/df=1.788, CFI=.935, TLI=.918, RMSEA=.067, SRMR=.061, but at higher complexity. However, near-null specific loadings, a Heywood anomaly, and weak general loadings render the bifactor model invalid. Balancing fit and parsimony, we retained the correlated three-factor structure as the final measurement model and interpreted scores at the level of the three dimensions. Multiple-group CFA evaluated the three-factor ISIS-21 across gender, residence, and professional domain. Using ΔCFI≤.01 and ΔRMSEA≤ .015 as decision rules, configural, metric, and scalar invariance were supported for all three grouping variables; strict invariance was met only for residence, and equality constraints on latent means did not degrade fit in that comparison. These results justify comparisons of latent means across gender, residence, and professional domain, while comparisons of observed scores should be limited to residence. The findings indicate that the validated instrument meets the necessary psychometric standards and may serve as a useful tool for evaluating spiritual intelligence among Romanian employees.