After the crude model of a protein is constructed we enter the stage of refinement. The parameters of the model are altered to improve the agreement between the model and the experimental observations. If we construct a function which reflects the discrepancy between the these two, refinement becomes the minimization of this function.
All are familiar with fitting models to data in this fashion. Finding the ``least-squares'' line through a collection of points is the classic example. However line fitting is easy and refinement is hard -- The difference lies in the relationship between the model and the experimental observations. This paper will discuss how the particulars of our structural model limits our ability to interpret diffraction data.