The auditory modality is fundamentally a temporal sense, requiring analysis of changes in sound signals on timescales ranging from microseconds to minutes. To generate a faithful representation of changes in sound intensity and frequency over time, sound offsets (disappearances) as well as sound onsets (appearances) must be encoded by the auditory system. Here we review the computational significance, perceptual roles, anatomical locations, and cellular and network origins of sound-offset responses in the mammalian auditory brain. We show that sound-offset responses arise from mechanisms and pathways distinct from those producing sound-onset responses, and are likely to be essential for auditory processing of temporally discontinuous sounds such as speech. 1 Sound-Offset Responses in the Auditory System Mechanoelectrical transduction mechanisms in the ear encode transient changes in sound pressure with sub-millisecond precision. The auditory nerve then provides to the brain a high-fidelity representation of sound frequency, intensity, and timing, usually including sharp increases in nerve activity following sound onsets and decreases following sound offsets [1]. Many neurons in the auditory brain also represent sound onsets and offsets in this way; however, others produce bursts of activity following sound offsets (e.g., [2]), or following both onsets and offsets (e.g., [3]). Historically, most studies of the central auditory system have ignored sound-offset responses, or dismissed them on neurobiological, acoustical, or perceptual grounds. Neurobiologically, sound-onset responses are much more prevalent in the auditory system than sound-offset responses [4], especially in anaesthetized animals [5], in which the majority of in vivo studies have been performed. Acoustically, offsets of natural sounds tend to be less abrupt than onsets [6], and sound offsets are often obscured by reverberation, especially in enclosed environments. Lastly, perceptually, sound offsets are less salient than sound onsets [4,7,8].