Multi-stable perception is a phenomenon characterized by spontaneous changes in the appearance of a stimulus despite unchanging visual input. A well-known example of multi-stable perception is Rubin’s ambiguous faces/vase stimulus which fluctuates between being seen as two profile faces or as a central vase shape. Previous work has shown that pre-stimulus BOLD activity in face-selective cortex is predictive of which interpretation will be perceived in the subsequent stimulus. Here, we aimed to further characterize the nature of this pre-stimulus neural activity using the high temporal resolution of EEG. On each trial the participants saw exactly the same ambiguous Rubin stimulus with inter-stimulus intervals ranging from 3-15 seconds (Poisson distributed). Trials were separated into those reportedly perceived as faces and those reported as vases. Fixation was monitored with an eyetracker to ensure no fixation differences (and thus no stimulus input differences) between these two sets of trials. As corroboration for participants’ subjective reports, we looked at the amplitude of the post-stimulus N170 ERP component. We found that the N170 was significantly larger for trials reported as faces than those reported as vases suggesting the participants’ subjective perceptions were, in fact, in line with their reports. To assess pre-stimulus activity, we computed the spectral power across 4-30 Hz in the second leading up to each stimulus. We found that theta and beta oscillations preceding the stimulus were predictive of the reported direction of figure-ground assignment. Lower power was observed for face percepts than vase percepts. These findings specify the nature of spontaneous pre-stimulus activity leading up to figure-ground organization. Furthermore, the results provide parameters (e.g. frequency and precise timing) which can be used to prime or stimulate the brain in future experiments to establish a causal link between pre-stimulus activity and perception.