(Originally posted: 6/27/2011)
After too long a break, I’m returning to the promised theme of thing about science that even scientists get wrong. This time, I’ll be discussing the relationship betwene theories and hypotheses. Full comments after the jump.
It’s common to read in elementary and high school science textbooks that a hypothesis is just an educated guess, and that if you test and confirm the hypothesis, it becomes a theory. Thanks to the woefully inadequate philosophy of science education at the higher levels of education these days (i.e., college), many people get at least through their bachelor’s degree without having those misconceptions corrected, and consequently you’ll sometimes hear even those with advanced degrees saying that theories are tested hypotheses.
The really insidious problem, of course, is that you can’t accept a theory without testing a hypothesis, and therefore on first glance, the relationship seems correct. However, for those who think through the logical relationships and try to give formal definitions to the concepts, it becomes clear that the theory comes first, gives you the hypothesis, and the two are really separate things. This realization came to me during graduate school (during ASU’s intense method and theory course), and I expect it comes to most scientists around the same time.
To really see the relationships, you need a formal definition of both concepts. A formal definition is one that first places the concept in a larger category, then differentiates it from other examples of the came category. Defining the term theory is probably one of the more controversial tasks in the philosophy of science, unfortunately. I tell my beginning anthropology students that, if you locked ten scientists in a room and told them they couldn’t come out until they’d agreed on a definition, in 24 hours, you’d have two survivors and three definitions. Nevertheless, here’s my crack at defining theory: A theory is an explanation of observed relationships among variables that invokes relationships with unobservable variables. This places the term in a larger category (“explanations of observed relationships among variables”), then explains what’s unique about it (that it “invokes relationships with unobservable variables”). Those unobservable variables are called theoretical constructs. They can only be observed indirectly, through their interactions with observables, and those observable relationships only make sense when the constructs are taken into account. In the physical sciences, they include things like “energy” and “molecules.” Some will undoubtedly shout out that molecules actually exist, and we’ve seen them! First, theoretical constructs must exist if the theory is correct–that’s the point of the theory–and second, no one has seen a molecule. They’ve seen, for example, a picture on the screen of a piece of lab equipment that they believe represents a molecule. This is indirect observation. In the social sciences, theoretical constructs include things like “power” or “meaning,” which are a little easier to understand as inherently unobservable.
So, by this definition, a theory is a variation on the idea that there is more to the world around us than meets the eye. The things we see are only part of the total picture, and we need to take into account the unseen in order to fully explain the world. There is nothing inherently wrong with a theory that posits, for example, tiny demons shuffling around molecules to explain chemistry. It is every bit as much a valid scientific theory as Dalton’s atomic theory. Both are simply attempts at explanation.
So how do we judge between them? This is where a hypothesis comes in. A hypothesis is a prediction of observable patterns that assumes a particular theory accurately describes the world. (Larger category: “predictions of observable patterns;” distinctive trait: “assumes a particular theory accurately describes the world.) It takes the form: IF <theory> is true, THEN <pattern> will be observed. Logically, one cannot generate a hypothesis before generating a theory. The theory comes first, then generates the hypothesis. The experimental process is the process of looking for that pattern.
Some basic high school algebra is necessary now. The hypothesis is of the structure IF A THEN B. Let’s say the experiment fails, and the predicted pattern is not observed. Assuming there were no flaws in the experimental design, we can conclude NOT(B), THEREFORE NOT(A). The theory (at least in its current form) is proven false, and we can further conclude that the theoretical constructs invoked in the theory do not exist. However, if the experiment is a success, and the pattern is observed, we cannot conclude the theory is absolutely true. IF A THEN B and B, THEREFORE A are not logically equivalent. For example, another contradictory theory C may also predict the pattern B. Simply observing B would therefore not differentiate between A and C as true theories. On the other hand, using a Bayesian approach to probability, we might say B, THEREFORE A IS MORE LIKELY. That is, P(A|B)>P(A), or, we are now more confident that A is true, but still not certain.
Regardless of this, there is no amount of testing that will turn a hypothesis into a theory, because each has a very distinct logical structure. Theories come first. The theory explains the world. To test the theory, you generate a hypothesis. The hypothesis tells you what to look for to test the theory. As might be a recurring theme in these “misconception” posts, the distinctions are probably a bit esoteric for general consumption, but scientists specialize in the esoteric, and would do well to keep these distinctions in mind!