A lot of the so-called systems composers have this thing that the system is always right. You don't fiddle with it at all. Well, I don't think that. I think the system is as right as you judge it to be. If for some reason you don't like a bit of it you must trust your intuition on that. I don't take a doctrinaire approach to systems.

As soon as you externalize an idea you see facets of it that weren't clear when it was just floating around in your head.

Another way of working is setting deliberate constraints that aren't musical ones - like saying, "Well, this piece is going to be three minutes and nineteen seconds long and it's going to have changes here, here and here, and there's going to be a convolution of events here, and there's going to be a very fast rhythm here with a very slow moving part over the top of it." Those are the sort of visual ideas that I can draw out on graph paper. I've done a lot of film music this way.

Once I started working with generative music in the 1970s, I was flirting with ideas of making a kind of endless music - not like a record that you'd put on, which would play for a while and finish.

I had a lot of trouble with engineers, because their whole background is learning from a functional point of view, and then learning how to perform that function.

Stop thinking about art works as objects, and start thinking about them as triggers for experiences.

I wanted to get rid of the element that had been considered essential in pop music: the voice.