Properties of the one-element model as applied to paired associate learning