Where is a bunch of frequencies composing a sound, such as a car horn or a baby cry, represented as a whole? While all the information about incoming sensory stimuli is present, at least implicitly, starting from the sensory organs and most probably in all stations of a sensory system, the nature of the sensory representation changes. In the auditory system, for example, sounds are initially decomposed into their frequency components. To study where the representation of sounds as wholes emerges in the auditory pathway, we used stimulus-specific adaptation (SSA) – the reduction in the responses of a neuron to a common sound (standard) which does not generalize to another, rare sound (deviant) – as a tool. We designed a new set of wideband stimuli, termed tone clouds, that have identical frequency components but are nevertheless perceptually distinct in order to neutralize frequency-specific mechanisms of SSA. Tone clouds evoked early and substantial SSA in primary auditory cortex (A1) but only late and minor SSA in the inferior colliculus (IC) and the medial geniculate body (MGB). These results imply that while in subcortical auditory stations sounds are largely represented in terms of their frequency components, in the auditory cortex they are already represented as abstract entities.