Spectrum sensing is the most important procedure for the operation of cognitive radio systems. To overcome the shortcomings of the conventional spectrum sensing techniques, considerable attention has been recently focused on the use of Modulation Classification (MC) for spectrum sensing. It relies on the fact that all primary users employ one modulation scheme or another for the transmission over the wireless channel; therefore, detecting any modulation scheme would be enough to confirm the presence of the primary user's signal in the wireless channel. This paper discusses the spectrum sensing for cognitive radio systems using modulation classification and provides a critical review of the existing methods; where gaps in the knowledge base are highlighted and directions for future research are suggested. © 2017. NSP Natural Sciences Publishing Cor.