Hypothetico-Deductivism

Later in the 17th century, a more plausible view of scientific methodology began to emerge in the work of Christiaan Huygens (1629 - 1695). His crucial idea eventually was clarified and championed by other scientists and philosophers including, notably, William Whewell—the same man who coined the term ‘scientist.’ Huygens, Whewell, and others suggested that scientific inquiry follows a special procedure:

1. The scientist begins by carefully observing and collecting data about some phenomenon (or range of phenomena).
2. Then, she formulates a hypothesis—a provisional theory—to explain the data she has collected.
3. Next, she logically deduces from the hypothesis a prediction about what to expect in some other situation that has not yet been observed.
4. Finally, she tests her hypothesis either by producing the relevant situation in an experiment or by finding the relevant situation in nature, and she observes whether the prediction was correct. If the prediction was wrong, the scientist discards that hypothesis and tries a new one; if the prediction was correct, she deduces another prediction and tests the hypothesis again.

This procedure has been dubbed the hypothetico-deductive (H-D) method because it involves deducing predictions from hypotheses. The H-D method is employed in nearly every field of scientific inquiry today, not only in the natural sciences but also in the social sciences. In fact, the method is so ubiquitous that it is (somewhat misleadingly) often called “the” scientific method, though it certainly isn’t the only method scientists employ. We’ll examine the H-D method more closely later in this chapter, where several specific variations or applications of the method will be considered. For now, let’s focus on the relevance of the H-D method to the demarcation problem. Could the H-D method itself be a distinctive feature of genuine science?

Though he pioneered the H-D method, Huygens was not especially concerned with demarcating science from other disciplines. (That’s understandable, since he was not only an innovative scientist but an accomplished mathematician and busy inventor who had little time for philosophical reflection about the nature of science as a whole.) William Whewell is another story. His coinage of the all-important word ‘scientist’ in 1834 was only a prelude to an ambitious refiguring of the philosophy of science two and a half decades later. Following Bacon’s vainglorious precedent of styling his own book as a replacement for Aristotle’s Organon, Whewell published a work titled Novum Organon Renovatum (the New Organon Renewed) in 1858. His view of science resembled Bacon’s in suggesting that hypotheses are formulated via inductive reasoning, but he emphasized the importance of testing hypotheses by deducing predictions from them.For further discussion of Whewell’s relation to later hypothetico-deductivist views, see historian Laura J. Snyder’s article on William Whewell in the Stanford Encyclopedia of Philosophy. Although Whewell is associated with hypothetico-deductivism, Snyder points out that he “explicitly rejected the hypothetico-deductive claim that hypotheses discovered by non-rational guesswork can be confirmed by consequentialist testing.”

Later philosophers of science, especially in the 20th century, dropped the inductive aspect of Whewell’s view entirely. They argued that the process by which a scientific hypothesis is formulated doesn’t matter. The hypothesis need not be the conclusion of an inductive inference. It might, instead, be the product of a scientist’s creative imagination. Or, just as well, it could be inspired by the scientist’s personal religious views. It could even come from a dream!That apparently happened, at least once: In the mid 1800’s, German chemist August Kekulé was trying to figure out the molecular structure of the organic compound benzene when he had a reverie or daydream of a snake biting its own tail. This led him to consider the hypothesis that the benezene molecule is circular, which his subsequent experiments confirmed to be true. (At least, Kekulé claimed this is how he arrived at the hypothesis. Historians aren’t certain whether he was serious or joking.) The origin of the hypothesis is unimportant. What matters is how the hypothesis is tested.

The formulation of scientific hypotheses, on this account, is a thoroughly subjective process. The testing of hypotheses, on the other hand, is supposed to be objective. So long as the scientists tests her hypothesis using the hypothetico-deductive method, she is doing science correctly. This view, called hypothetico-deductivism, provided demarcation criteria that seemed to rescue the objectivity and special epistemic status of science. According to hypothetico-deductivism, a theory or hypothesis is objectively confirmed to the extent that it yields accurate predictions, and it is falsified (shown to be false) when its predictions do not accord with the new data. Thus, genuine science exhibits two distinctive features:

1. Scientific inquiry always employs the hypothetico-deductive method.
2. The formulation of scientific hypotheses may be a subjective process, shaped by the personal biases and agendas of a scientist. Nevertheless, the process of deducing predictions and testing them is a rigorous, objective procedure. Scientific hypotheses are confirmed or falsified by testing their predictions in a fully objective way. This objectivity is what distinguishes genuine science from pseudoscience.

Before examining objections to this demarcation proposal, it’s important to distinguish hypothetico-deductivism from the hypothetico-deductive method. The former is a philosophical view about science; the latter is a common method of science. The philosophical view called hypothetico-deductivism regards the H-D method not as just one of many scientific methods, but as the one and only unique, distinctive method of all science.

The strong claim that science always employs the H-D method falls into an obvious objection: it’s just not true that this is the only method scientists ever use. The H-D method isn’t even the primary method of inquiry in some branches of science. For example, the science of paleontology—the study of fossils—doesn’t primarily involve deducing predictions and testing them. Granted, paleontologists use the H-D method on occasion. Sometimes, they deduce predictions from evolutionary hypotheses and test them by searching for predicted features in the fossil record.Evolutionary creationist Darrel Falk offers the following example: “If evolution is true, there would be a time when there would have been a lineage from fish to land animals (tetrapods). According to dating mechanisms, we know that land animals first appeared on earth about 370 million years ago. Until 2004, no good transitional species had been found. Neil Shubin and his collaborators hypothesized that evolutionary theory predicts that if they search hard enough in rocks a little older than 370 million years of age, they will find fossils of species that have both fishlike and tetrapodlike characteristics. They knew that there were very few rock formations on earth of just the right age but identified one set in northern Canada. Beginning in 1999, they put together a crew that searched the rock formations for fossils that had the expected characteristics. Over a five-year period, the group went out to that area every summer and screened very carefully for fossils. Over the first four summers, they found nothing. Then toward the end of the summer of 2004, which they had decided would be the last year of the search project, they found a nearly complete specimen of a species that had exactly the predicted characteristics—a fossil which had certain fish and certain tetrapod characteristics.” - Todd Charles Wood and Darrel R. Falk, The Fool and the Heretic: How Two Scientists Moved beyond Labels to a Christian Dialogue about Creation and Evolution (Grand Rapids: Zondervan, 2019), 130. However, most research in paleontology involves other methods, such as inference to the best explanation. (The latter method involves comparing rival hypotheses to determine which provides a better explanation for observed patterns in the fossil record. We’ll revisit it later in this chapter.) See also the chapter on Non-deductive Inferences in my Skillful Reasoning e-book for further discussion of inference to the best explanation and its relation to other modes of reasoning. So, the first criterion above is too restrictive: it can’t be a necessary condition for science, because genuine science doesn’t always use the H-D method.

Moreover, using the H-D can’t be a sufficient condition for science either. Some non-scientific fields of inquiry—such as history, philosophy, and even mathematics—occasionally make use of the H-D method, or methods closely analogous to it. For example, a mathematician might hypothesize that some equation is a theorem in a given mathematical system, then deduce some consequence from her conjecture to check whether it holds. Similarly, a historian might hypothesize that Marco Polo confused Madagascar with Mogadishu, then make predictions about what kinds of descriptions are likely to turn up in the explorer’s memoirs if he mistook one location for the other. (Incidentally, many historians think that is exactly what happened.) Of course, we could say that the mathematician and the historian really are doing science when they use the H-D method, but that would defeat the purpose of having a demarcation criterion in the first place. We wanted to know what is unique about science and how it differs from other fields of inquiry. In light of the aforementioned examples, it seems that using the H-D method is neither necessary nor sufficient for doing science.

Further problems arise for the hypothetico-deductivist’s claim that the objectivity of this special method is what distinguishes genuine science from pseudoscience. According to criterion 2, above, scientific hypotheses are supposed to be confirmed by successful predictions (and falsified by failed predictions) in a fully objective way. However, the confirmation of a scientific theory cannot be fully objective, for several reasons:

1. First, even if a theory yields successful predictions, it doesn’t logically follow that the theory is true. (To say otherwise would be to commit the logical fallacy of affirming the consequent.To see the problem, consider the following argument: If you overslept, you’ll be late for class. You’ll be late for class. Therefore, you overslept. This argument is fallacious: even if the premises are both true, the conclusion may be false. You might be late for other reasons. Perhaps you’ll be late because the subway train was delayed, or because of a traffic jam, or because you simply lost track of time. Although the hypothesis that you overslept would indeed explain your late arrival, it is not the only possible explanation. For a formal definition of the fallacy of affirming the consequent, see this page of my Skillful Reasoning e-book. (If you’re unfamiliar with the logic symbols, you’ll want to start reading at the beginning of the chapter on Propositional Logic.)) It might follow that the theory is more likely to be true, but it’s difficult to say exactly how much more likely, if at all. For reasons discussed later in this chapter, the degree of confirmation a theory receives from its successful predictions appears to be at least partly subjective, dependent on the presuppositions and biases of the individual scientist or of the scientific community.

To appreciate the force of this first point, remember the underdetermination problem: no matter how much data we have, there are always infinitely many possible theories compatible with the data. These theories can’t all be true. They can’t all be likely to be true, either. So, why should the one theory under consideration be regarded as probably true when other theories might yield equally successful predictions? Privileging one’s own theory in this way sounds like a subjective bias!

Of course, there are cases where one hypothesis seems obviously superior to its rivals. Consider again the “Alternative Hypotheses” illustration on the previous page. The ellipse hypothesis is simpler than the wobbly alternative, and it is tempting to think that the evidence objectively supports the simplest hypothesis. But defining ‘simplicity’ in a purely objective way turns out to be about as difficult as defining ‘science.’ Even if we could come up with purely objective criteria to determine which hypothesis is the simplest, moreover, we still face a difficult question: why should we regard the simplest hypothesis as the one that’s most likely to be true? Isn’t that a bias or presupposition? We have no guarantee that the universe is simple, and it’s far from obvious whether we have any fully objective reason for favoring simpler hypotheses.

Additionally, there are two other ways in which theory confirmation seems to rely on assumptions or presuppositions:

2. Most scientific theories are universal in the sense that they are meant include all instances of some type of object or phenomenon. (For example, atomic theory applies to all atoms in the universe.) Moreover, there is no limit to how many predictions can be deduced from a theory. No matter how many predictions we test, therefore, we will have tested exactly zero percent of the theory’s predictions! Is it reasonable to conclude that a theory is probably true when we have tested 0% of its predictions? If we regard the theory as true—or even just likely to be true—it may seem as though we are committing the fallacy of hasty generalization. To mitigate this worry, we must presuppose that nature is uniform, i.e., that the laws and regularities we experience here in our tiny corner of the universe are representative of the whole. That’s not a trivial assumption, and it might not be justifiable with strictly objective considerations.
3. Throughout the history of science, well-confirmed theories are repeatedly superseded by new theories. Thereafter, the old theories are seen as false. For example, Aristotle’s theory of gravity was superseded by Newton’s theory, which in turn was superseded by Einstein’s theory. Similarly, Thomson’s model of atoms was replaced by Rutherford’s model, which was replaced by Bohr’s model, which was replaced (or, at least, refined) by quantum mechanical models. Each new theory casts its predecessors as mistaken. Therefore, some philosophers of science argue, we should regard even our best contemporary theories as probably false, and we should expect them to be superseded by new theories in the future. (This argument has been called the pessimistic meta-induction.The pessimistic aspect of the argument is obvious enough, but the term “meta-induction” calls for explanation. The argument was deemed a meta-induction because scientific reasoning traditionally was considered inductive reasoning (as in Bacon’s view), and this argument makes an inductive inference based on all of those supposedly inductive inferences of outdated science.) A similar worry can be raised against the idea that theory confirmation is fully objective: it sounds strange to say that all of those refuted theories had been objectively confirmed, in the past, by the old data. More plausibly, the apparent confirmation these defunct theories had enjoyed was partly a function of subjective factors, including background beliefs scientists held at the time.

These objections to hypothetico-deductivism do not imply that science is merely subjective, of course. Later in this chapter, we’ll examine the issue of scientific objectivity more carefully. As we’ll see, scientific reasoning—like all human reasoning—is complex, and it involves both objective and subjective elements. (I’ll also try to clarify precisely what I mean by “objective” and “subjective” in this context.) For now, I want to highlight a common feature of the above three challenges to scientific objectivity: all of these objections concern the confirmation of theories or hypotheses. But what about the falsification of hypotheses? Could the refutation of a theory, at least, be a straightforwardly objective matter? We’ll consider that question on the next page.