Auditory masking is more than that. The psychoacoustic model used for MP3s will discard some frequencies during others, like you said, but it will also discard

Auditory masking is more than that. The psychoacoustic model used for MP3s will discard some frequencies during others, like you said, but it will also discard "information" proceeding and following an acoustic event. While not a perfect analog, it is similar to how vision works. If you are in a bright room and the lights go out, it takes you a little time to readjust to the darker room before you can see things again. While your irises were adjusting, the chair and take were still there, and light reflecting from those objects were still reaching your eyes, but you were unable to see them. It works the other way too. If you are still in that dark room and the lights come on, it will take your eyes a little time to readjust, less time then it took your eyes to adjust to the dark, but still your eyes are effectively discarding that information. The psychoacoustic model in MP3s is even more bizarre. It turns out that some frequencies, when heard BEFORE another frequency, your brain will throw that first frequency away. It is unintuitive, but that has been proven in laboratory settings. Knowing how the majority of humans discard the same auditory information under different conditions, MP3s are compressed by throwing away the same information that the brain would normally discard. The Fourier transform is a significant part, but it isn't the whole story.