One of the points that creationist Ken Ham made in his debate with Bill Nye, and presumably is still making on his site "Answers in Genesis," is that we have to distinguish between experimental and historical sciences. According to his argument, physics is an experimental science, evolution and geology are historical. Since the first type produces testable knowledge, and the second doesn’t, we can safely disregard the second and substitute Creationism in its place, without doing any damage to the first. I suppose a lot of people are inclined to dismiss this point as just obviously fatuous because Ken Ham made it, and because he’s attacking science, and therefore has to be resisted on all fronts. The instinct is sound, but as a strict matter of logic it doesn’t follow, and it’s not really rational to lurch to one extreme simply because one’s opponents have taken up residence on the other.
Maybe there’s an opportunity to learn about science here, so I’d like to take up Ken Ham’s point. He’s right that there is a distinction to be drawn, although he’s wrong to conclude, on that basis, that the results of the historical sciences can be disregarded at will. More on this later. The main difference between the experimental and the historical sciences is that the former are concerned with the elucidation of regularities across all events but specific to none of them, while the latter are concerned with the elucidation of specific, unique, non-repeatable events.
To take a familiar example, suppose that you wanted to study the trajectory of artillery shells. And suppose, further, that you didn’t know the specific equations that govern their flight, but had to rediscover them. A sensible way to go about this would be to set up an artillery piece, fire it a hundred or a thousand times (preferably under tightly controlled circumstances), and record the results. That is, you would conduct an experiment. The goal of the experiment would not be to describe the trajectory of any one shell, but rather of all the shells, and on that basis to isolate the underlying regularity, which might be called a “law” or a “physical constant.” Once you had isolated this, you could go on to make predictions about shells fired under less tightly controlled circumstances, and under those circumstances (say, a misfire, or a strong breeze) the behavior of the particular shells would change. The underlying regularity, however, would not.
Thus a “law of nature,” usually describable through an equation, does not refer to what must happen in any one particular case, but rather to a statistical regularity that pertains, we believe, across all cases. We can never be completely certain that we've got the statistical relationship right because we can’t observe all instances of the given phenomenon—we can observe many, and we can make reasonable inferences on that basis, but the universe is very large and very old and we only have significant experience of a very tiny corner of it, and over a very short period. Who knows, maybe the constants will change tomorrow. It’s not likely, but it’s at least conceivable (the question of how we know that physical constants do not change is a special case of a broader problem in philosophy, the problem of induction. For more on this, see Russell’s Chicken or Nelson Goodman’s concept of “Grue.”)
Suppose, however, that you had a different question in mind. Imagine that in the course of your experiments one of the shells misfired, the artillery piece exploded, and your ballistics lab burned to the ground. You might, naturally, be curious about the reason for this, and decide to conduct an investigation. Supposing that you had vast time and resources at your disposal, it would be possible to repeat the earlier, experimental approach, i.e. to build thousands of ballistics labs and fire tens of thousands of artillery pieces hundreds of thousands of times, all under tightly controlled circumstances, wait for some of them to burn down, look for statistical correlations, and then try to discover the underlying law that governs the burning down of ballistic laboratories.
Obviously, this wouldn’t be a very practical approach, but more to the point, it wouldn’t actually answer your question, which was not “Why do ballistics labs burn to the ground?” but why did your ballistic lab burn to the ground. It might be helpful to know what the underlying statistical regularity is—that, for instance 21% of burned-down laboratories are the result of an improperly installed fire-suppression system, 16% of faulty lab equipment, 11% of operator error, and so on, and that might point you in a useful direction—but it still wouldn’t tell you which of those factors was relevant to the destruction of your particular lab. What you would need to do is pick over the debris (the way they do when an airliner crashes), consult cell-phone records (if they're available), talk to the survivors (if there were any), and so on.
But no matter how thorough your investigation, you could never be entirely certain that you had got to the bottom of things. If you discovered, for instance, that the lab assistant was inexperienced, had been struggling with alcohol addiction, and was upset over a fight with his girlfriend, you might suspect that operator error played a role in the demise of your laboratory. But this would really only be circumstantial evidence. It’s also possible that, though inexperienced and struggling with difficult circumstances, your assistant had performed his tasks correctly, and that the problem was actually a poorly manufactured shell, a gas leak, or something else. Again, it would be a case of drawing inferences, and this would necessarily be somewhat indirect, because the particular event of your lab burning down on that particular day will never recur, and therefore cannot be directly observed. It is forever in the past, where we cannot go.
None of this is to say that experiments can tell us nothing about past events, or that particular events are irrelevant to establishing law-like regularities. As we have seen, even absurd experiments would have some value in the case of burning laboratories, and an experiment just is a collection of particular events. We can’t insist on an absolute distinction in this area any more than we can with the ship of Theseus. Rather, it’s a matter of emphasis, of choosing the best method for the best task.
Nevertheless, the person who insists that the experimental method is very sound and ought to be applied where possible certainly has a point. The precision of physics is largely due to the power of its methodology, and no doubt many researchers wish they had something of equal power at their disposal, even if there are reasons why experimentation in particular is not adequate. In this respect the results of the historical sciences are generally less certain than those of the experimental. The reason is that a law is a sort of if-then statement, of the form “under conditions X, the likely result will be Y.”
Thus a ballistics expert can tell you what will likely happen if you fire a shell at such and such an angle and velocity under such and such conditions, but not that you have fired it, or will fire it. The law is constant, but the event is contingent. In the same way, when discussing particular events, the question of whether those events actually occurred is distinct from the question of what laws would have governed them if they had. If knowledge of the trajectory of one artillery shell was all you had, you would never be able to deduce the underlying regularity that had governed its flight, nor, knowing only that regularity, would you be able to say whether a shell had been fired at a particular time and place. Rather, we derive our knowledge of the regularity from many repetitions of similar instances, and it is because we can repeat them so frequently and under so many different circumstances that we can have a high degree of confidence in the abstractions we arrive at in this way.
It is important to recognize, however, that there is no reverse procedure for establishing that a non-repeatable event has in fact occurred in the past, no matter how confident we are of the corresponding regularities. We have to look at the evidence available to us in the present and, for lack of a better word, guess. It just doesn't produce the same kind of confidence that the experimental procedure does because there's no good way to eliminate wrong interpretations. A proposition in physics suggests an experiment that might refute it. It faces, as Karl Popper would say, a severe test. A proposition in a historical science like evolutionary biology or geology faces the somewhat milder test of correlating the data. That is, it has to provide a satisfying explanation for the evidence, but since different people find different things satisfying, this isn’t quite so rigorous. The best we can do in this case is immerse ourselves in the appropriate body of evidence and evaluate it fairly (that is, without foregone conclusions), stay open to new approaches and cognizant of our own predispositions, and discuss our findings honestly. It remains the case, however, that we are dealing with interpretations.
Now, that doesn't mean all interpretations are equally valid. To return to Ken Ham, his interpretation is not a very good one because it doesn’t even pass the rather modest threshold outlined above. I don’t doubt his honesty, but he doesn’t seem to have a very good grasp on the evidence, and the view that the earth is about 6,000 years old simply makes nonsense out of the geological record, as well as radiometric dating methods on which our current estimation of the age of the earth rests. He’s clearly not evaluating the evidence without foregone conclusions, nor is he open to new approaches, since it is rather the point of his argument that he already knows that his particular reading of Genesis is the only right one. Thus on his approach, evidence that might disconfirm his theory is itself disconfirmed for that very reason. The conclusion is being used to evaluate the evidence, where the rational procedure would be to use the evidence to evaluate the conclusion. It’s as if our ballistics expert simply disregarded all the shell trajectories that didn’t correspond to his prior idea of what the underlying regularities were, or as if our fire investigator disregarded the evidence that a gas leak had caused the explosion because he already knew that it was the lab assistant’s fault. We might reasonably expect to derive a flawed theory from this flawed method, and indeed this has proven to be the case with Creationism. The problem is with the underlying disposition that it represents, or in other words with a refusal to take the evidence seriously.
Daniel Halverson is a graduate student studying the History of Science and Technology of nineteenth-century Germany. He is also a regular contributor to the PEL Facebook page.