ON MODE AND IMPLICATION IN APPROXIMATE REASONING