Notice the in-ear monitors, and the microphones. The mikes are facing the front of his guitar, and he's hearing himself and a click track through the headphones. Each additional layer is recorded while the existing layers are playing back through the headphones. When he sang, he could hear the guitar part he recorded earlier. When I recorded a bass or cello part, I heard the vocals and guitar.
Each track is independent, and the volume can be changed without affecting anything else. For example, when I recorded a bass track, I set my live bass part louder than what was being played back (to ensure that I could hear myself), with the vocals at about half the volume of the guitar. Then a quick reset of the sliders, and everything is back where it was. Pretty nifty, eh?
