actually works surprisingly well for many phenomena, including the photoelectric effect and stimulated emission. But it fails for three crucial predictions: