When analyzing complex acoustic scenes, the brain continuously groups related sounds and segregates others to form coherent perceptual streams. Neural mechanisms of streaming have been proposed based on spectral differences, and more recently also relying on f0 differences with unresolved spectral fine structure. These data suggest that feature-specific forward suppression is subserving streaming based on a variety of perceptual cues, and results in explicit neural representations of auditory streams within auditory cortex. Using functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG), the present study investigated the human cortical activity during streaming based on spatial cues, and compared it to the cortical activity elicited by streaming based on f0 differences. 12 listeners were presented with 32-s sequences of a repeating ABBB pattern composed of 125-ms harmonic complex tones (lowpass filtered at 5kHz) at a normalized sound intensity level of 70 dB SPL. The fundamental frequency (f0) of the A tones was always 180Hz (i.e. the lower harmonics of the tone complex were spectrally resolved), and the tones were lateralized to the left by an inter-aural time difference (ITD) of -687.5µs. Three conditions were compared: (1) a control (no-streaming) condition, where the parameters of the B tones were identical to the A tones, (2) a spatial-streaming condition, where the B tones were lateralized to the right by an ITD of 687.5µs, yielding an ITD difference of 1375µs between A and B (no f0 difference), and (3) an f0-based-streaming condition, where the f0 of the B tones was 120.1Hz, such that the f0 difference between A and B equaled 7 semitones (no ITD difference). Psychophysical responses, recorded during the measurements, revealed reliable streaming in the spatial as well as the f0-based condition. The cortical activity measured by fMRI and MEG was stronger when the listeners perceived the tones segregated into two perceptual streams for the spatial (ITD-based), as well as the f0-based condition. In fMRI, the activation time course during both streaming conditions exhibited a similar shape and increased amplitude, this increase being slightly higher for f0-based streaming. Throughout auditory cortex, both streaming conditions involved topographically almost identical areas, with the f0-condition producing somewhat more expanded activation. In MEG, the amplitudes of the P1m, N1m, and P2m components evoked by A tones were significantly larger when listeners perceived streaming. The increase was generally similar, yet showing interactions between components. These results support the hypothesis of explicit neural representations of streaming in auditory cortex, served by feature-specific forward suppression. In particular, the data add evidence for spatial cues in streaming and show that these spatial cues involve the same sites as f0 and spectral cues for streaming. Research supported by DFG grant GU593/3–1