Paired-comparison wagers were offered to Ss under 2 different methods of encoding decision-parameter information. Method 1 provided S with a set of 4 elements of nonindependent parameter information, Method 2 with a subset comprised of 2 independent elements. Ss' choices showed significantly closer correspondence to an EV-maximization decision rule under Method 2, for both independent groups and repeated-measures designs. With repeated measures, it was found that given sufficient decision time, Method 2 transferred positively to Method 1, with the opposite true (negative transfer) when order of administration was reversed. It was concluded that the basic problem for S in choice situations is the selection of decision rules rather than of decision alternatives. Decision rules may then be asymmetrically transferrable across different encoding methods.