According to the object-based account of attention, multiple objects coexist in short-term memory (STM), and we can selectively attend to a particular object of interest. Although there is evidence that attention can be directed to visual object representations, the assumption that attention can be oriented to sound object representations has yet to be validated. Here, we used a delayed match-to-sample task to examine whether orienting attention to sound object representations influences change detection within auditory scenes consisting of 3 concurrent sounds, each occurring at a different location. On some trials, the 2 scenes were identical; in the remaining trials, the locations of 2 sounds were switched. In a control experiment, we first identified auditory scenes, in which the 3 sounds were unambiguously segregated, for the subsequent experiments. In 2 experiments, we showed that orienting attention to a sound object representation during memory retention (via a retro-cue) enhanced performance relative to uncued trials, up to 4 s of memory retention. Our study shows that complex auditory scenes composed of cooccurring sound sources are quickly parsed into sound object representations--which are then available for top-down selective attention. Here, we demonstrate that attention can be guided toward 1 of those representations, thereby attenuating change deafness. Furthermore, the effects of retro-cues in audition extend analogous findings in the visual domain, thereby suggesting that orienting attention to an object within visual or auditory STM may follow similar processing principles.