The recent interest in the imaging possibilities of photonic crystals (superlensing, superprism, optical mirages, etc.) call for a detailed analysis of beam propagation inside a finite periodic structure. An answer to the following question was sought: "Where does the beam emerge?" We found that, contrary to common knowledge, it is not always true that the shift of a beam is given by the normal to the dispersion curve. This phenomenon can be explained in terms of evanescent waves and a renormalized diagram yields the correct direction.