: Rock fracture toughness, a fundamental parameter in rock fracture mechanics, describes the intrinsic resistance of rock against initiation and propagation of macroscopic fractures. This parameter is used in a wide variety of applications in different fields of rock mechanics. Thus, measuring accurate and precise value of this parameter is very important. However, measurement method(s) of this parameter is still a matter of debate in the literature. On the other hand, rock properties in microscale are always leading factors in physical/mechanical behavior of rock. In many cases this relation is properly understood but so far, enough insight about how grain size distribution affects fracture toughness is not provided. Although experimental tests have a significant role in rock fracture toughness studies, numerical techniques in numerous instances show more advantages and capabilities. However, scanty attempts concerning use of numerical methods in this field are reported. In the present research, via more than 140 experimental tests, fracture toughness of three marbles (i.e. Neiriz, Baghat and Maron), which are different only in grain size distribution, are measured under mode I, mixed mode I-II and mode II loadings using both Hollow Centre Cracked Disc (HCCD) and Cracked Chevron-Notched Brazilian Disc (CCNBD) testing specimens. Quantification of the marbles grain size distributions is achieved directly from their thin sections. Based on determined grain size distributions, numerical samples of the marbles are modeled using Distinct Element Method (DEM) so that, macroscopic behaviors from numerical triaxial compressive tests are in agreement with the responses from the tests in laboratory. Afterward, using these numerical samples, fracture toughness tests in the experimental part of the research are simulated. Independent from the testing specimen, comparison between the marbles results shows that mode I fracture toughness ( K IC ) is correlated inversely with grain size. For each marble, HCCD specimen, in comparison with CCNBD, yields lower values of fracture toughness. This difference is negligible in pure mode I loading; however, becomes larger as loading condition transits from mode I to mode II. So that, pure mode II fracture toughness ( K IIC ) values from the two specimens are not comparable. Among measured physical/mechanical parameters, Brazilian tensile strength ( ? tB ), Schmidt hammer hardness and P-wave velocity ( V P ) are correlated with K IC . In numerical pure mode II fracture
