Analogical transfer in 3- to 5-year-olds was examined in three studies where the children were required to notice the common underlying goal structure of a set of problems. The children were either required to recall the prototype story before tackling the transfer problem, or were explicitly prompted to attend to the common goal structure. Subjects who spontaneously focused on the goal structure in their recall, or who were prompted to do so, transferred efficiently regardless of age. Children who did not represent the problems at the level of underlying goal paths, but instead attended to interesting surface features of particular stories, failed to transfer. Children as young as 3 years of age have the underlying competence to transfer a common problem solution; level of representation rather than age determines transfer efficiency. Transfer flexibility is not a simple function of age but depends on the level of analysis afforded the base analogy. The results are discussed in terms of emergent theories of mental models for learning via analogy.