The paper describes an experimental and numerical study of size effect on concrete cylindrical specimens in splitting tensile test. Own experimental campaign was performed on specimens with 5 various diameters from D = 74, 105, 150, 192 and 250 mm with hardboard loading strips (distributed load according to standard methods) scaled proportionally to the specimen diameter. The crack opening-control system was applied to obtain the post-peak behaviour of all tested specimens including catastrophic behaviour (snap-back). The tested specimens at a certain point were unloaded and scanned with novel high-resolution micro tomography to analyse the macro cracks and phenomena like aggregate breakage, crack branching etc. at the aggregate level. Based on realistic mesostructure the discrete element method (DEM) 2D model of 3 specimens with diameters of D = 74, 150 and 250 mm were constructed and tested. The fracture was analysed at macro and micro-level in DEM and directly compared with microCT scans. DEM simulations revealed additional information related to the loss of material strength and ductility with increasing specimen size (size effect). The simulation and experimental results were in good agreement. © 2022 The Authors