The water example showcases the germ of the idea pretty well. Stimuli don't need to be in 1-1 or even linear correspondence with physiological responses to those stimuli. Or maybe I'm misunderstanding his argument.