Ten years ago, animal models of noise-induced hearing loss predicted three cortical neural correlates of tinnitus resulting from noise-induced hearing loss: increased spontaneous firing rates, increased neural synchrony, and reorganization of tonotopic maps. Salicylate also induces tinnitus, however, the cortical correlates were reduced spontaneous firing rates, unchanged neural synchrony but some change to the tonotopic map. In both conditions increased central gain, potentially a correlate of hyperacusis, was found. Behavioral animal models suggested that tinnitus occurred, albeit not in all cases. The study of the neural substrates of tinnitus in humans is currently strongly based on network connectivity using either spontaneous EEG or MEG. Brain imaging combined with powerful analyses is now able to provide in excellent detail the lay out of tonotopic maps, and has shown that in people with tinnitus (and clinical normal hearing up to 8 kHz) no changes in tonotopic maps need to occur, dispensing therefore of one of the postulated neural correlates. Patients with hyperacusis and tinnitus showed increased gain, as measured using fMRI, from brainstem to cortex, whereas patients with tinnitus without hyperacusis only showed this in auditory cortex. This suggested that top down mechanisms are also needed. The open problems can be formulated by the following questions. 1) Are the neural substrates of tinnitus etiology dependent? 2) Can animal results based on single unit and local field potentials be validated in humans? 3) Can sufficient vs. necessary neural substrates for tinnitus be established. 4) What is the role of attention and stress in engraining tinnitus in memory?