A method is disclosed that breaks the "one line, one location" paradigm of teleconferencing in the prior art. The teleconference bridge in the illustrative embodiment is able to utilize more than one audio channel from each location, where there are multiple signal sources present in the room. As a result, the bridge is able to determine acoustically whether two are more endpoints are collocated with each another. During an initialization sequence, the bridge transmits special audio signals to one or more endpoints present in a particular sound field; those endpoints then play the signals out of their loudspeakers. Based on a characteristic (e.g., amount of correlation, signal strength, etc. ) of the signals received at each microphone present in the same sound field, the bridge determines whether to include or exclude signals that are received from a first endpoint when preparing a signal for transmission to a second endpoint during a conference call.