Multiple- versus single-problem training and variations of solution rules in the formation of learning sets.