Experiments explored the influence of amplitude modulation and spatial separation on detectability of a broadband noise target masked by an independent broadband noise. Thresholds were measured for all combinations of six spatial configurations of target and masker and five modulation conditions. Masker level was either fixed ͑Experiment 1͒ or roved between intervals within a trial to reduce the utility of overall intensity as a cue ͑Experiment 2͒. After accounting for acoustic changes, thresholds depended on whether a target and a masker were colocated or spatially separated, but not on the exact spatial configuration. Moreover, spatial unmasking exceeded that predicted by better-ear acoustics only when modulation cues for detection were weak. Roving increased the colocated but not the spatially separated thresholds, resulting in an increase in spatial release from masking. Differences in both how performance changed over time and the influence of spatial separation support the idea that the cues underlying performance depend on the modulation characteristics of the target and masker. Analysis suggests that detection is based on overall intensity when target and masker modulation and spatial cues are the same, on spatial attributes when sources are separated and modulation provides no target glimpses, and on modulation discrimination in the remaining conditions.