I just finished reading "Detection Theory, A User's Guide" authored by Hautus, Macmillan, & Creelman. I wholeheartedly recommend this book to any scientist working on psychophysics, perception -- or in fact psychology at large as the concepts of Signal Detection Theory are often overlooked by some subfields of social psychology or psycholinguistics.
One very interesting aspect that I never fully realized before is how Signal Detection Theory allows us to explore the very essence of our internal representations using only external (behavioral) measurements. In particular: do we represent the world using continuous internal values or binary (all-or-nothing) categories as in the case of the high-threshold theory?