Five experiments explored the utility of subvocal rehearsal, and of an inner-ear/inner-voice partnership, in tasks of auditory imagery. In three tasks (reinterpreting ambiguous auditory images, parsing meaningful letter strings, scanning familiar melodies) subjects relied on a partnership between the inner ear and inner voice, one similar to the phonological loop system described in the short-term memory literature. Apparently subjects subvocally rehearsed the imagery material, which placed the material in a phonological store that allowed the imagery judgement. In a fourth task (distinguishing voiced and unvoiced consonants in imagery), subjects still subvocally rehearsed, but seemed to need no additional phonological store to respond correctly. In this case they may have consulted articulatory or kinesthetic cues instead. In a fifth experiment (making homophone judgements), subjects hardly even needed to subvocally rehearse, a result suggesting that homophone judgements rely on some direct route from print to phonology. We consider the breadth of the partnership between the inner ear and inner voice, the level that subvocal rehearsal occupies in the cognitive system, and the functional neuroanatomy of the phonological loop system.