Although ejection fraction (EF) is the cornerstone of the assessment of left ventricular (LV) systolic function, its measurement faces a number of challenges related to image quality, assumptions of LV geometry, and expertise. The aim of this study was to test the inter-observer variability of EF and GLS measurement in patients with a broad spectrum of LV function, between physicians and investigators (Inv) with different levels of expertise. In 122 patients, EF and GLS were measured by 4 Inv blinded to each other with different level of experience in echocardiography; EF was measured using 3 methods: visual assessment, biplane Simpson's method, and auto-EF method. GLS was measured from the 3 apical views. A significant difference for LVEF and for LVGLS was considered to be >10 and >2 absolute values, respectively. Intra-observer agreement was excellent for visually assessed EF (ICC = 0.87, P < .001) and GLS (ICC = 0.82, P < .001) and good for EF measured by Simpson's method (ICC = 0.70, P < .001) and auto-EF (ICC = 0.72, P < .001). Intra-observer and inter-observer agreements were excellent for GLS with ICCs above 0.8. GLS discordance between the 4 Inv was not significant. Discordance in EF and GLS measurements among the Inv was not related to image quality or wall motion abnormalities. Although EF has proved its prognostic value in various cardiovascular entities, GLS seems to be more reliable for serial assessment of LV function, demonstrating lower intra- and inter-observer variability, even by different physicians with variant level of expertise.