The field of emotional intelligence (EI) research has yielded two rather distinct ways of conceptualizing and measuring the construct, which also differ greatly in terms of their scientific value: mixed models of EI, although commercially successful, prove inadequate when subject to scientific scrutiny; by contrast, the Mayer and Salovey ability model of EI meets most of the proposed criteria for establishing a scientifically meaningful EI construct. Its current operationalization, the MSCEIT, has thus far been found to exhibit good reliability, as well as convergent-dicriminant and structural validity. The present study examined the psychometric properties of the Serbian version of the test in a sample of 250 high school graduates. Our results show that the reliabilities of Serbian MSCEIT scores are equivalent to those reported for the American standardization sample; more specifically, reliability coefficients are high for the two Area scores and the Total score (r≥ .86). The Total EI score shows low to moderate correlations with standard measures of academic intelligence (r=.244 - .429; p<.01), and very low correlations with the Big Five personality traits (r<.25) - a pattern that is regarded optimal in establishing convergent-discriminant validity. The factor structure of the Serbian MSCEIT closely corresponds to that of the original test: the theoretically proposed 1-, 2- and 4-factor solutions all prove empirically sustainable; however, in the 2- and 4-factor solutions, the structure of certain EI branches diverges from that which is theoretically expected and incorporated in the scoring system. The obtained data certainly recommend the Serbian MSCEIT for standardization and practical use, in which case they should be supplemented with data regarding the test’s predictive validity. On a more basic level, our results suggest the possibility of assessing a scientifically meaningful EI construct, defined as the ability to reason with „emotional“ information.