Previously, we proposed a 'dielectrophoretically activated cell sorting' (DACS) system, which can be applied to the separation of various cells without labeling them with magnetic or fluorescent materials. By using a specific frequency range, it was verified that the dielectric material properties of cells can be exploited in order to collect target cells from a cell mixture. However, because about 20% of the error in separation efficiencies was always detected, it became a subject to investigate. Therefore, we assumed that variance in cell size might be responsible for the errors. By measuring the diameter of human breast cancer cells (MCF 7) and classifying separation efficiencies according to diameter, it was verified that the errors in separation efficiency are affected by cell size. Moreover, the existence of a 'blind spot,' i.e., a space with weak force within the dielectrophoretic field generated by a pair of electrodes, was proven by commercial code CFD-ACE +® . With simulation and experimental results, we demonstrated that some efficiency errors occur because small sized cells (between 15 and 20 µm) pass through the 'blind spot'.