It is widely understood that the 'difficulty' of a particular computaiton varies according to how the input data are presented. With the data presented one way, achieving particular computational effects may require sophisticated and elaborate operations. With the data presented differently, it may be possible to achieve the same effects much more straightforwardly. There is thus a tradeoff between computation and representation. What is not widely understood is the impact this tradeoff has within learning and cognition in general. This paper argues that in more advanced forms of learning the tradeoff is a major consideration affecting the way in which particular learning tasks can be accomplished. Examples are given that demonstrate this effect and its wider epistemological and developmental implications are discussed.
This paper is not available online