CHAPTER EIGHT
Scientific Methodology
What is the foundation of all conclusions from experience? This implies a new question, which may be of more difficult solutions and explication.
—David Hume (1748/1974)
Many of us learned in grade school a rather simple and stilted process labeled the scientific method. A science education website aimed at schoolchildren still proposes such a formulation in six steps:
1. Ask a question
2. Do background research
3. Construct a hypothesis
4. Test your hypothesis by doing an experiment
5. Analyze your data and draw a conclusion
6. Communicate your results (Steps of the Scientific Method, 2009)
For schoolchildren, this may be a good introduction, but it does not begin to hint at the controversies that the presumptions underlying this “method” have sparked. This method has been challenged both in theory and in practice. Philosophers of science have pointed to theoretical problems underlying this method; and historians of science (and some scientists themselves—see Bauer, 1992) have challenged the presumption that science is always, normally, ideally, or even ever performed according to this rubric. We begin with the most fundamental problem underlying this method and continue with various attempts to resolve this problem.
THE PROBLEM OF INDUCTION
An inductive generalization is the logical process of attributing a quality to a whole class of objects based on our limited experience of some sample of that class. The interesting, and perhaps troublesome, aspect of this logical process is that it is common not just in scientific methodology but in daily life as well. We each make these types of inferences every day, often without even noting or realizing it. Merely opening your front door requires what seems to be an implicit inductive inference that turning the knob has opened the door in the past and will do so today. Scientific claims, laws, and theories are largely based on inductive inferences. Isaac Newton’s law of inertia was inductively inferred from the observation of many individual objects that manifested this quality. From the continual confirmation of the hypothesis that all matter has the quality of inertia, the inference is made that the claim of inertia as a property of all matter is in fact a law about matter. In this example, we see the direct correlation between induction and empiricism. Empiricism holds that knowledge is attained through empirical observation; observation of the eyes, ears, and so forth. Empirical observation is always observation of particulars. That is, we observe particular dogs, not the species dog or the concept of dog. We observe particular instances of matter manifesting inertia, not matter itself or inertia itself as general concepts or realities. Logical positivists/empiricists, being rather strict empiricists, noted this aspect of observation and saw at least part of the method of science to take these particular instances and generalize from them inductive inferences as a product of knowledge. The particular observations in themselves are not real or strong knowledge. The inferences based on these observations are bolder, more useful forms of knowledge. The claim that a particular billiard ball has the property of inertia is a less bold and less useful claim than the claim that all matter has this property. The claim that a new antibiotic will kill a certain range of bacteria in one patient is a less bold and less useful claim than the claim that this antibiotic will kill such bacteria in all patients. However, the stronger claim in both instances is more difficult to establish and much riskier in terms of being knowledge. As we learned in Chapter 4, inductive reasoning only leads to probabilistically supported conclusions. Thus, all inductively reached scientific claims, laws, and theories are true only to a degree of probability and always carry with them the logical possibility of being proven false. Yet there is an even deeper problem underlying this limitation. According to the classical empiricist David Hume (a person clearly admired by both logical positivists and logical empiricists), induction is not only limited by probability but is not even “founded on reasoning, or any form of understanding” (Hume, 1748/1974, p. 328).
Let us take a simple example of inductive reasoning. Most people accept, without giving it much thought, that the sun will rise tomorrow.1 Why do they believe this? The simplest explanation is that the sun has risen every day in their lives and in recorded history in the past. So, the inference might be expressed in the following simple argument:
The sun has risen every day in the past.
Therefore, the sun will rise tomorrow.
At first blush this may seem like a clear argument. Yet, there is an important premise missing:
The sun has risen every day in the past.
The future will continue to be like the past.
Therefore, the sun will rise tomorrow.
This new version provides a general principle (the italicized premise) that better connects the original premise with the conclusion, more clearly connecting the past to the future. But this new claim is clearly open to doubt and in need of evidentiary support. Being a general principle, it is not something known through direct experience, as explained previously, but reached through some form of inference or as a self-evident, analytic claim. The denial of an analytic claim is a self-contradiction: for example, “All bachelors are unmarried.” The denial of that claim would be “It is not the case that all bachelors are unmarried,” or more clearly, “Some bachelors are married.” But to claim that any bachelor is married is a contradiction of the concept of “bachelor.” However, the denial of the claim “The future will continue to be like the past” does not result in a self-contradictory claim. To say that the future will not continue to be like the past is an empirical and speculative claim. It can only be proven true by observing the future. So, this principle is not an analytic claim and cannot be said to be true on that basis. Thus, it must be an empirical claim and require empirical support. One might attempt to provide support for this claim by adding the premise “In the past the future has been like the past.” Yet to use this claim to support the “The future will continue to be like the past” claim would be to employ once again the type of reasoning we are trying to prove—to assume once again that the future will be like the past by again being like the past. This is a type of logical error (fallacious reasoning) philosophers refer to as circular reasoning or begging the question: employing what is intended to be proven in order to prove it.
CAUSALITY
Another form of inductive reasoning, called causality, is subject to the same type of criticism. Indeed, a causal inference may be understood as a type of inductive generalization. When one makes a causal claim, which is common in science, one asserts that some event A causes some second event B. This is an empirical, not an analytic, claim. For once again to deny it does not result in a contradiction. For example, one might assert that HIV causes AIDS. To deny this claim is only contradictory if we presume the definition of HIV as that which causes AIDS, but that again results in a question-begging argument. Indeed, if this were an analytic claim, it would not have been so difficult to reach. The early history of AIDS research shows the proposal of many possible causes of AIDS: amyl nitrate, a weakening of the immune system by a succession of (especially sexually transmitted) infectious diseases, and so forth. And the fact that HIV as the cause of AIDS is so widely accepted and so strongly supported by research now that today those who will deny this claim are almost seen as maintaining a contradiction, they in fact are not. As small a possibility that this causal claim is false that there is, there still is that possibility. It is a claim that can only be justified through experience—in this case experience in terms of scientific observation and research—and inductive logic. In saying that A causes B, we are referring to two conceptually separable events. When we analytically say that all bachelors are unmarried, the two concepts (bachelors and being unmarried) are not conceptually separable because the concept of “being unmarried” is included in the concept of “bachelor.” To say that HIV causes AIDS is again to synthetically connect these conceptually distinct events or phenomena. For a simpler example, let us say the cue ball strikes the eight ball and causes the eight ball to move. Here again we have two conceptually distinct events: the cue ball striking the eight ball and the eight ball moving. As they can be held distinct in this manner, they are not necessarily or analytically connected. To connect them we must appeal to experience and something known as inductive inference.
The deeper problem, according to Hume, has to do with our concept of causality itself. This concept includes three primary elements. The first is temporality. That is, when we assert that A causes B, it is presumed that A temporally precedes B. You never have an effect before a cause. The second element is spatial contiguity. The cause and the effect must come into physical contact. We say the cue ball causes the eight ball to move because it strikes the eight ball. We do not attribute the movement of the eight ball to the five ball, which is resting still at the other end of the table. Even when we seem to attribute causality at a distance, we don’t really (setting aside quantum mechanics for the moment). To say that pressing the power button on the remote control causes the television (TV) to turn on leaves unsaid but implied that pressing the power button causes an infrared beam, which, when operating on a sensor on the TV, causes the TV to turn on. When we say the remote power button causes the TV to turn on, there is implied a causal chain. But beyond these two elements, there is a third, which is more conceptually troublesome. Saying that A causes B means more than that A precedes B (temporality) and that A and B occur in the same place (spatial contiguity). Also part of our concept of causality is that there is some power in A to bring about B. Hume referred to this element as the necessary connection. The trouble with this element is that while the first two (temporality and spatial contiguity) are empirical, this necessary connection is not. A necessary connection implies that any causal claim is analytical, but as we have already seen, it is not. This necessary connection is something that we cannot see or perceive in any manner. “Causality” is, itself, beyond perception. Any causal inference we make can only be based on the former two elements: repeated observations of A occurring before B, and A and B occurring at the same place. Thus, we inductively generalize from particular observations of these two elements to the claim that the two events are connected by this seemingly metaphysical concept of causality. Yet, Hume doubted that such an inference is rationally supported:
The bread, which I formerly eat, nourished me; that is, a body of such sensible qualities was … endued with such secret powers: but does it follow, that other bread must also nourish me at another time, and that like sensible qualities must always be attended with like secret powers? (Hume, 1748/1974, p. 329)
As with the sun rising tomorrow there is no logical reason to accept that bread will continue to nourish me, or that the cue ball striking the eight ball will in the future cause movement in the eight ball, that HIV will continue to result in AIDS. Hume ultimately could find nothing more than custom or habit to attribute our inductive inferences to. That is, from viewing two events together again and again, we by habit or custom come to expect them to be conjoined in the future: a new day conjoined with a rising sun, HIV conjoined with AIDS, the striking of the eight ball conjoined with movement of that ball, the eating of bread conjoined to nourishment. But this expectation is merely psychological, not logical.
Now Hume did not completely reject causality or induction in general. As a human being he had to accept and work with those concepts on a daily basis, as we all do. His main problem was that as a philosopher, especially one interested in understanding and justifying scientific knowledge, he could find no logical justification for induction. And therein lay the problem: as a practical human being or as a scientist he has to employ and accept the practice of induction. But as a philosopher he can find no reasoning to back it up, leaving the status of inductively “inferred” claims as knowledge fundamentally uncertain.
HYPOTHETICO-DEDUCTIVISM
Empiricists in the 19th century (e.g., John Stuart Mill) and in the 20th century (logical positivists/empiricists) worked to find a better inductive methodology. In much of the logical positivist/empiricist work you can find many hopeful statements about the improvements on induction to come. A reliance on induction follows logically from the central logical positivist doctrine of verifiability. Many philosophers proposed strategies to solve the problem of induction. Although none was successful in addressing the fundamental problem, a number of these strategies were valuable for highlighting problems attendant to the problem of induction and possibly making the problem of induction seem less intractable.
One simple strategy is the broadening of the role of the hypothesis. If we look back to the inductive method at the beginning of modern science, in the work of Francis Bacon (1561–1626), we see a very simple approach in his Novum Organum. He outlined his empirical investigation into the concept and phenomenon of heat. His method included merely listing everything that had the properties of heat (fire, the sun, etc.) and those things that lacked heat (ice, cold earth, etc.), then abstracting from all the hot things what they had in common, and which those things on the list of non-hot things lacked, to determine what caused heat. There is a simple practical problem to this method that reflects the more general problem of induction. A list of hot things and a list of non-hot things would each be at least indefinite if not infinite in length.2 The scientific method of hypothetico-deductivism, as employed by the logical positivists/empiricists, like Carl Hempel (1966/2000), was intended to address this problem. The basic structure of the method is to suggest a likely hypothesis to explain an event or predict the recurrence of an event, then logically deduce what would and would not follow logically (deductively) if that hypothesis were true—thus, “hypothetico-deductivism.” Experiments then are structured to see if the logically deduced consequences follow. The introduction of a hypothesis provides some boundaries and guidelines, thereby limiting the observations that would be relevant to include in an investigation, for, in the words of Hempel, “a collection of all the facts would have to await the end of the world” (1966/2000, p. 45). But what exactly does “relevant” mean? What makes a fact or an observation relevant to an investigation? First, it means relevant to the hypothesis in question. A little more deeply it means that “either its occurrence or its nonoccurrence can be inferred from” the hypothesis in question. Hempel employs the example of physician Ignaz Semmelweis, who, while working in the 1840s at Vienna General Hospital, investigated the cause of childbed fever, which took the lives of many new mothers in Vienna. What was particularly interesting and concerning was that there was much more childbed fever in the First Maternity Division than in the Second Maternity Division. One hypothesis he investigated was that the birthing position might be the cause. Women in the Second Division more commonly birthed laterally, and those in the First Division more commonly birthed on their backs. It seemed that those delivering on their sides had less incidence of childbed fever. Thus, the hypothesis was that birthing in a supine position causes childbed fever. The question of relevance then hinges on what facts would be logically implied by this hypothesis. Most obviously a fact such as a rise or decrease in childbed fever on changing the birthing position would be logically relevant. That is, if the hypothesis were true, we would expect a change in the First Division to lateral birthing to decrease the incidence of childbed fever. If the hypothesis were not true, we would expect no change in changing the birthing position in the First Division. In fact, he found that changing the birthing position had no effect on the incidence of childbed fever and concluded then that the hypothesis was false. The logical structure of such an investigation would be as follows (where H is some hypothesis and E is a logically relevant event or fact):
If H is true, then E would follow (is true).
E does not follow (is not true).
Therefore, H is not true.
This is a logically valid argument. The form is known as modus tollens. Here is a simple example to illustrate the validity of the form in general:
If Bill is in Philadelphia, then Bill is in Pennsylvania.
Bill is not in Pennsylvania.
Therefore, Bill is not in Philadelphia.
Here it should be clear that the conclusion follows logically from the premises. If one is not in Pennsylvania, then one cannot be in Philadelphia—given that we are speaking of the specific Philadelphia that is in Pennsylvania, as the first premise asserts. The previous argument has the same modus tollens form and thus is deductively valid also. So it seems that childbed fever is not caused by birthing position. Another hypothesis Semmelweis investigated was that “cadaveric matter” was the cause. Most of the women in the Second Division were attended by midwives, while most of the women in the First Division were attended by physicians and medical students who would often come directly from performing autopsies. To test this hypothesis he had physicians and medical students wash their hands before moving from autopsies to birthing. This may seem like a ridiculously obvious solution, but bear in mind that in the 1840s the germ theory of disease was still a contentious, not widely accepted theory. And, as we might expect, on instituting a hand washing policy the occurrence of childbed fever decreased. Semmelweis thus concluded that childbed fever was caused by an infection delivered from cadavers by physicians and medical students to new mothers. The formal structure of the reasoning is a bit different in this case:
If H is true, then E would follow (is true).
E does follow (is true).
Therefore, H is true.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
