Work is currently underway to devise learning methods which are better able to transfer knowledge from one task to another. The process of knowledge transfer is usually viewed as logically separate from the inductive procedures of ordinary learning. However, this paper argues that this `seperatist' view leads to a number of conceptual difficulties. It offers a task analysis which situates the transfer process *inside* a generalised inductive protocol. It argues that transfer should be viewed as a subprocess within induction and not as an independent procedure for transporting knowledge between learning trials.
Download compressed postscript file