Binaural hearing aids potentially provide binaural cues that can improve the dectability and the spatial separation of multiple sound sources. This paper considers the use of binaural cues and the resultant spatial percepts on listeners' ability to separate simultaneous sound sources. In backgrounds with continuous noise or multiple talkers, the main problem is the detection of individual acoustic components. On the other hand, if a single masking sound is very similar to the target, and both target and masker are spectro-temporally sparse, as is the case with speech, the main problem, at least for listeners with normal hearing, is to decide whether a particular spectro-temporal feature belongs to the target source and to track that source across time. Although the subjective location of a sound source can help in grouping features across time, its effect is most easily observed in the absence of other differences between the sound sources.