Defenders of methodological naturalism often invoke definitional or "demarcation criteria" that say that all science must be observable, testable, falsifiable, predictive, and repeatable. Most philosophers of science now dismiss these criteria because there are too many exceptions to the rules they establish in the actual practice of science. Not all science involves observable entities or repeatable phenomena, for example --you can't watch all causes at work or witness all events happen again and again, yet you can still make inferences about what caused unique or singular events based on the evidence available to you. Historical sciences such as archeology, geology, forensics, and evolutionary biology all infer causal events in the past to explain the occurrence of other events or to explain the evidence we have left behind in the present. For such inference to work, the cause invoked must now be known to produce the effect in question. It's no good proposing flying squirrels as the cause of the Grand Canyon, or a silt deposit as the cause of the Pyramids. Squirrels don't dig giant canyons or even small ones, and silt doesn't move heavy stone blocks into an ordered three-dimensional array. However, we know from our experience that erosion by running water can and does produce gullies, then arroyos, and by extension, canyons. We know that intelligent agents have the necessary design capabilities to envision and build a pyramid. No natural force does. These are inferences based on our present knowledge of cause and effect or "causes now in operation." The theory of intelligent design also qualifies as historical science. We cannot directly observe the cause of the origin of life or repeat the events we study in the history of life, but we can infer what cause is most likely to be responsible, as Stephen Meyer likes to say, "from our repeated and uniform experience." In our experience the only thing capable of causing the origin of digital code or functional information or causal circularity is intelligence and we know that the origin of life and the origin of animal life, for example, required the production of just such things in living systems. Even though other demarcation criteria for distinguishing science from non-science are no longer considered normative for all branches of science, it is worth checking to see how well intelligent design fares using criteria that are relevant for an historical science. Briefly, although the designing agent posited by the theory of intelligent design is not directly observable (as most causal entities posited by historical scientists are not), the theory is testable and makes many discriminating predictions. Steve Meyer's book Signature in the Cell, Chapters 18 and 19 and Appendix A, discusses this thoroughly.