I think the most reliable way to advance a scientific field is to find a technique used in a different field and apply it to yours. Even more reliable if you use techniques from theoretical math.