Tungsten-based memristors possess many advantages as candidates for memristive devices, including gradual changes in resistance states and memorization and learning functions. However, most previous reports mainly focus on studying synaptic learning rules instead of analysing the internal mechanism that results in the exterior learning rules. Herein, we discuss stacked Au/WTiOx/Au and Ti/WTiOx/Au devices in which the function of the resistance switch is realized by the externally induced local migration of oxygen ions. The consecutively adjustable multilevel resistance of the Au/WTiOx/Au device may be due to the variation in the barrier width and height in high oxygen vacancy concentrations. Additionally, the high and low resistance states of Ti/WTiOx/Au devices are considered as a result of the connection and rupture of the conductive filaments at low concentrations of oxygen vacancies. The physical mechanism construction and state-full synapse development through the control of ion migration provide insight into the applications of oxide-based memristors in neuromorphic computation.