I guess they mean that they just have a summary of the sound, not enough to listen to it or anything. It would be a bit like just having the hash of a piece of data, not the data itself.