<p>Understanding the size effect on the shear behavior of backfill–rock interfaces is essential for reliable evaluation of underground backfilled structure stability. This study investigates how specimen size influences shear behavior through a combination of laboratory experiments and multi-method 3D discrete element simulations. Four interface construction approaches (affine, splice, cutting, and scaling) were developed to represent in-situ rough rock surfaces. Laboratory direct shear tests were performed on scaling specimens to calibrate micro-mechanical parameters for the DEM simulations. Simulation results reveal that under the affine, splice, and cutting methods, increasing specimen size leads to a significant reduction in peak shear stress, a transition of failure mode from axial compression to interface-localized shearing, and a reorientation of internal force chains toward the shear direction. In contrast, the scaling method shows nearly constant shear strength and failure patterns across sizes. Morphological analysis indicates that traditional roughness descriptors (maximum height difference, height variance, and 3D roughness) exhibit weak correlation with the observed size effect. Instead, the progressively decreasing height–width ratio in the affine, splice, and cutting methods is identified as the dominant factor. This is further supported by the vertical distribution of micro-cracks, which increasingly localize along the interface in larger specimens. The findings highlight the critical role of geometric proportions in governing scale-dependent interface behavior and offer a mechanistic basis for accurately predicting the shear performance of large-scale backfill–rock systems.</p>