Can you falsify an earth science hypothesis?
Answer: No. Read on for why.
Science, they tell us in grade school and middle school and high school and often into college, has four simple steps: Observe, formulate a hypothesis, test that hypothesis, reject or accept that hypotheses (usually testing against a null). In this model, hypotheses should be falsifiable. Is this done in earth science?
Consider the big mountain building hypothesis of the late 19th and early 20th century: geosynclinal theory. The idea was that parts of the earth’s crust would warp down, accumulate a lot of sediment, heat up and get injected with magma and deformed by thrust faults and rise up to be mountains. One side was continental (the miogeosyncline) and the other we’d now consider to be oceanic (the eugeosyncline). Lots of embarrassing problems emerged: among them, fossils from one part of the geosyncline were totally different from those elsewhere in the geosyncline, so magical barriers had to be made (various flavors of eugeanticlines). Arguably many of these demonstrated that the theory was wrong, but instead scientists proposed ways that the theory could still be right. It isn’t until plate tectonics overcame the objections to its predecessor of continental drift that geosynclinal theory was abandoned. It took a theory to kill a theory.
Consider another case: the origin of the southern Rocky Mountains (aka the Laramide orogeny). One proposal was that friction from a flat subducting slab on the base of North America created enough stress within the plate to make it buckle and break far to the east of the plate boundary. This is actually a nice, physically elegant way to make mountains in the interior of a continent without demolishing everything between those mountains and the plate boundary. A 1988 paper by Peter Bird incorporated this physics in a model of the Laramide and, as a good hypothesis should do, it predicted that mantle lithosphere older than the Laramide should be missing west of Colorado. When it was shown that such lithosphere remains, the theory died, right? No; it continues to this day as the preferred theory for the Laramide.
Why would this be? Is this the gang that couldn’t hypothesize straight? Two things are going on: one is an absence of a null hypothesis, the other is testing hypotheses against a complex system out of your control.
Lab experiments have controls. Feed these plants fertilizer and don’t give it to these other plants. The null hypothesis would be that they will grow the same; if they don’t then fertilizer affects their growth. But a lot of earth science simply lacks a null hypothesis. What is the null hypothesis for the flat slab model for the Laramide? If you figure it out, please let GG know. There are certainly some aspects of earth science where you can manage null hypotheses, but in historical contexts, they are very rare.
The other problem is that the earth is a very complex place. No reductionist theory will ever explain everything, so you will always have observations that will be at odds in some way. To take a simple example, Steno’s Law of superposition holds that in a stack of sedimentary rock, the younger will lie over the older rock, and such rocks are often dated with fossils. But there are instances where you may find something that looks wrong: there are places where you will find Cretaceous (say 70 million years old) fossils sitting in sediment above Miocene (~12 Ma) rock. Does this falsify Steno? No: in this case, erosion of Cretaceous rock nearby freed up Cretaceous fossils that were incorporated into the younger sediment. Obviously you could promote this as a test of Steno (if Steno is right, there should be evidence for the greater youth of the higher rock layer) but typically, after many other tests of Steno, we choose to say, yeah, those Cretaceous fossils are probably reworked and move on. [This aspect of earth science is why we as a community usually don’t bother with a lot of things like “this exposure disproves radiometric dating,” much to the dismay of some strident voices; chasing the oddities at the end of a Gaussian distribution aren’t going to yield much insight].
So if you advance a hypothesis and some observations disagree with it, the hypothesis might live on. But when do you admit the hypothesis isn’t so useful? In essence, the null hypothesis in geology is “we don’t have any idea.” Can you get by with that?
These days, the answer is no. To get funded, one must (almost) always have a testable hypothesis and you say “if we find X, this confirms the hypothesis, if not, it refutes it.” But in practice, you are usually contrasting two competing hypotheses, mainly for the reasons outlined above. In fact, if you don’t have a competing hypothesis, you generally don’t get funded.
So the geological practice of “multiple working hypotheses” as espoused by Chamberlain years ago may be more than just a fruitful way to guide a field geologist to examine all evidence, it might be a necessity to make any progress at all. Basically, if there is one hypothesis out there, especially if it bears on multiple fields (as, for instance, the flat slab hypothesis does), then the presentation of new results will generally be interpreted within that framework, even when it requires invoking special explanations for some of the new observations. The reason is that the authors will perceive that there is lot of other stuff they’d have to disprove to claim the hypothesis was wrong. So, for instance, gravity anomalies in the Sierra Nevada were interpreted for a long time in terms of a thick crust even as gravity gradients in places demanded an alternative explanation; it wasn’t until the thick crust was shown not to exist that the door was opened to a greater spectrum of possibilities.
So the earth scientific method might well be observe, develop two or more viable hypotheses, determine a situation where they make conflicting predictions, observe/experiment, and refine.