Reading music is something a child can learn, and once understood, it becomes such a natural process that it is no longer a conscious effort. If we were to dissect this `natural process,' we might hypothesise that reading music is decomposed into two parts: the visual recognition of graphical shapes; and the application of our musical knowledge to derive its meaning. A computer paradigm that models this structure would be a vision system connected to a knowledge base. Imagine an Optical Music Recognition (OMR) system where the user describes the simple graphical shapes found in music using a customised drawing package, and expresses the musical knowledge necessary to correctly interpret these simple graphical shapes, using a specially designed musical language. Such a system would capture the essence of reading music, forming a versatile foundation.