In recent years the overlap between visual perception and memory has shed light on our understanding of both. When ambiguous images that normally cause perception to waver unpredictably are presented briefly with intervening blank periods, perception tends to freeze, locking into one interpretation. This indicates that there is a form of memory storage across the blank interval. This memory trace codes low-level characteristics of the stored stimulus. Although a trace is evident after a single perceptual instance, the trace builds over many separate stimulus presentations, indicating a flexible, variable-length time-course. This memory shares important characteristics with priming by non-ambiguous stimuli. Computational models now provide a framework to interpret many empirical observations.