Concrete specimens of different sizes and shapes were made with various reactive aggregates and stored under conditions favorable to the development of alkali-silica reactivity (ASR), with their expansion measured with time along the three directions. They have been cast vertically (cylinders and prisms) or horizontally (prisms and larger blocks), using a vibrating table, a vibrating needle, or rodding. The expansion due to ASR was always greater in the direction perpendicular to the casting plane. The higher the number of flat and elongated particles in the reactive aggregate, the higher the coefficient of anisotropy, defined as the ratio between the expansions perpendicular and parallel to the casting plane. This coefficient was constant through the course of the expansion. It was generally higher for the cylinders than for the prisms, and still less for larger blocks. Consolidation by rodding induced anisotropy coefficients distinctly smaller than consolidation using a vibrating table, while a vibrating needle induced intermediate values; however, all methods gave constant volumetric expansion at least up to an important expansion level. For prisms cast horizontally and measured axially in accordance with the concrete test CSA A23.2-14A or ASTM C 1293, consolidation using rodding induced long-term (axial) expansions greater by 71% compared with consolidation using a vibrating table. In order to reduce the experimental variability of the test, only one method of consolidation should be allowed. When evaluating field concrete affected by ASR, it appears important to consider the orientation with respect to the casting plane of the core samples subjected to mechanical or residual expansion tests.