My research question sort of breaks down in to three different components. The first is timbre recognition. If the robot is going to learn how to play by listening to you, it is going to need to figure out which timbres you are actually using. Number 2 is timbre production, so again, if the robot is going to learn how to *use* timbres, that implies it is actually capable of producing a suitable variety of timbres. The first two items really just deal with isolated instances of musical sound, so the third item is timbral context, which looks at how you then combine these individual timbres within the context of actual music. So, how is the human using these timbres, and how can the robot use that to inform its own playing. I can't tell you about everything in my dissertation in one hour so I'm going to focus on these last two things...