We can recognize the melody of a familiar song when it is played on different musical instruments. Similarly, an animal must be able to recognize a warning call whether the caller has a high-pitched female or a lower-pitched male voice, and whether they are sitting in a tree to the left or right. This type of perceptual invariance to "nuisance" parameters comes easily to listeners, but it is unknown whether or how such robust representations of sounds are formed at the level of sensory cortex. In this study, we investigate whether neurons in both core and belt areas of ferret auditory cortex can robustly represent the pitch, formant frequencies, or azimuthal location of artificial vowel sounds while the other two attributes vary. We found that the spike rates of the majority of cortical neurons that are driven by artificial vowels carry robust representations of these features, but the most informative temporal response windows differ from neuron to neuron and across five auditory cortical fields. Furthermore, individual neurons can represent multiple features of sounds unambiguously by independently modulating their spike rates within distinct time windows. Such multiplexing may be critical to identifying sounds that vary along more than one perceptual dimension. Finally, we observed that formant information is encoded in cortex earlier than pitch information, and we show that this time course matches ferrets' behavioral reaction time differences on a change detection task.