How can selection and drawing support learning from dynamic visualizations ?