Whilst the metaphor of ‘fitness landscapes’ is widely applied in the evolutionary algorithm (EA) community, there are several assumptions requiring its application that are often ignored, such as the underlying structure of the search space and the ontological status of the values depicted by the landscape. By differentiating between valuation and evaluation surfaces, and surfaces of selective value, it is possible to show how each may be transformed by learning (and learning costs) and the inheritance of traits acquired therefrom. In doing so, two styles of learning are identified, rank respectful and rank transforming, and these are shown to behave differently under proportional and rank based selection schemes.