The Royal Road problem has been something of a thorn in the side of Genetic Algorithm (GA) research since it was first introduced in the early 1990s. This problem was designed specifically to be “GA-easy ” and to highlight the sort of problem that the GA promised to solve more efficiently than techniques such as hill-climbing. In practice, however, the Royal Road problem is not solved efficiently by the canonical GA and analysis of the problem has demonstrated significant weaknesses in the GA. This thesis extends the analysis of the Royal Road problem by moving beyond the issues of building block destruction and combination and looking at the more fundamental difficulty that the GA experiences: construction of new building blocks during the course of a run. It is demonstrated that existing techniques for improving building block supply are, at best, inefficient and fragile, and at worst require significant a priori knowledge of the structure of the problem. Mathematical analysis of the commonly used genetic operators shows that