Electrical capacitance tomography, based on the neural network multicriteria optimization image reconstruction technique, developed by this research group, is used to examine the column size effect on the real time, cross-sectional flow variation during the choking transition in a circulating fluidized bed. Two sizes of columns, 0.05 and 0.1 m i.d., are employed in this study for two types of particles, sand particles (group B) and fluid catalytic cracking catalysts (group A). For group B particles, the formation of the square-nosed slugs (0.05 m i.d. column) or the wall slugs (0.1 m i.d. column) is observed to occur in the regime transition when the gas velocity U-g is below the transport gas velocity U-tr or when the solids circulation rate G(s) is below the transport solids velocity G(s,tr). When U-g > U-tr or G(s) > G(s,tr,) the formation of open slugs for both 0.05 and 0.1 m i.d. columns is observed in the regime transition. Clearly, for group B particles, these regime transitions accompanied by a distinct bed structure change mark the choking transition. For group A particles, when Ug < U-tr or G(s) < G(s,tr,) a distinct change in the bed structure in the columns is also observed as the flow transits from the dilute regime to the turbulent regime, marking the choking transition. When U-g > U-tr or G(s) > G(s,tr,) however, the transition from the dilute regime to the dense regime is fuzzy. Even though this transition is noted in the literature as the choking transition, it involves little change in the bed structure.