A powder sandwich technique was applied to determine thoron (220Rn) and radon (222Rn) emanation rates for a granular material. The feature of this technique is the sample preparation, in which a granular material is put and fixed between two membrane filters. Airflow is directly given to this sandwich sample, will include thoron and radon emanated from the material, and then is transferred to the detector. This method makes sure that radon as well as thoron emanated is not retained in pore space within the sample volume, which is crucial for the appropriate emanation test. This technique was first introduced by Kanse et al. (2013) with the intention to measure the emanation of thoron-but not of radon-from materials having much higher 224Ra activity than 226Ra. In the present study, the methodology for the discriminative determination of thoron and radon emanation rates from a granular material has been examined using a flow-through scintillation cell and sandwich sample. The mathematical model was developed to differentiate total alpha counts into thoron-and radon-associated counts. With a sample of uranium ore, this model was experimentally validated by comparison between the scintillation cell and a reference detector that can discriminatively measure thoron and radon concentrations. Furthermore, the detection limits and uncertainties were evaluated to discuss the characteristics of this method. Key parameters for their improvement were found to be the background radon concentration and the leakage of radon from the measurement system, respectively. It was concluded that the present method is advantageous to a sample that has much higher 226Ra activity than 224Ra if the emanation fractions are similar between thoron and radon.