Raise mining is applied to extract raw materials in an underground mine. In this method stopes with cylindrical geometry are extracted and adjacent stopes are separated by a continuous pillar system. This pillar compound system is the result of the cylindrical stope geometry and differs from “classical” pillar systems. In contrast, the "classic" pillar systems comprise isolated and detached pillars. As pillars in a pillar compound system are restrained in all directions, it is expected that they behave differently compared to isolated, free-standing pillars. Currently, there is little to no experience regarding the behavior of such pillar compound systems. Thus, a laboratory test method has been developed and tests have been conducted. The test method takes real mine conditions into account by using similarity laws. Tests are uniaxial and both the pre- and the post-failure area are examined. In order to simulate the restraining effect of surrounding rock mass a test frame is built. The aim of the laboratory tests is to analyze and describe the behavior and the strength of the pillar compound system under different loading directions. The description is mainly about the strength, rigidity and fracture formation. In addition to loading directions, extraction ratio (size of the pillars) and pillar geometry (stope arrangement) are varied. Available numerical simulations describe the stress redistribution in the prevailing pillar compound system. They are used additionally to investigate the results of conducted laboratory tests. Results show a significant difference in the behavior and the strength of pillar compound systems under different loading directions. With vertical loading of the pillar compound system, no significant deterioration in strength due to extraction of the stopes is noted. According to test results, the lateral load of the pillar compound system turns out to be much more critical. In this case overstressing of the pillar compound system can occur at significantly lower loads. Based on these studies, it seems appropriate to backfill the excavated stopes. Thus, the need for more extensive experiments with backfill is outlined and suggested. The interpretation of the results has shown that there is room for improvement in data acquisition. Particularly the assessment of stresses in the pillars of the model turned out to be difficult.