Human and nonhuman primates have a remarkable ability to recall, maintain and manipulate visual images in the absence of external sensory stimulation. Evidence from lesion, single-unit neurophysiological and neuroimaging studies shows that these visual working memory processes are consistently associated with sustained activity in object-selective inferior temporal neurons. Furthermore, results from these studies suggest that mnemonic activity in the inferior temporal cortex is, in turn, supported by top-down inputs from multimodal regions in prefrontal and medial temporal cortex, and under some circumstances, from the hippocampus.