Distinct anatomical and functional pathways are postulated for analysing a sound's object-related (‘what’) and space-related (‘where’) information. It remains unresolved to which extent distinct or overlapping neural resources subserve specific object-related dimensions (i.e. who is speaking and what is being said can both be derived from the same acoustic input). To address this issue, we recorded high-density auditory evoked potentials (AEPs) while participants selectively attended and discriminated sounds according to their pitch, speaker identity, uttered syllable (‘what’ dimensions) or their location (‘where’). Sound acoustics were held constant across blocks; the only manipulation involved the sound dimension that participants had to attend to. The task-relevant dimension was varied across blocks. AEPs from healthy participants were analysed within an electrical neuroimaging framework to differentiate modulations in response strength from modulations in response topography; the latter of which forcibly follow from changes in the configuration of underlying sources. There were no behavioural differences in discrimination of sounds across the 4 feature dimensions. As early as 90ms post-stimulus onset, AEP topographies differed across ‘what’ conditions, supporting a functional sub-segregation within the auditory ‘what’ pathway. This study characterises the spatio-temporal dynamics of segregated, yet parallel, processing of multiple sound object-related feature dimensions when selective attention is directed to them.