6a. Explain the hypothetico-deductive method and its implications for testing scientific hypothesis
The hypothetico-deductive method (HD Method) is one of the main hypothesis testing methods across scientific disciplines. The method consists in the following steps:
There are some implications for using the hypothetico-deductive method. For example, it requires the scientific hypothesis to be testable. Only empirical hypotheses are testable – no hypotheses about immaterial souls or divinities are testable. In addition, the method does not guarantee the results. Scientific knowledge is always probable, never definitive. The best the method can do, therefore, is offer a strong likelihood that the prediction is correct. So, we need to consider alternative hypotheses. Finally, while a prediction may fail, the theory may still be correct. So, disconfirmation does not disqualify the theory. What we need to do, in that case, is consider auxiliary (additional) hypotheses.
To review, see The Hypothetical-Deductive Method.
Occam's Razor, also known as the principle of parsimony, is a rule that asserts the simplest explanation is the best. In the sciences, the rule admonishes unnecessary complications. The practical implications of this rule are not always positive. For example, in biology, where things can get "messy", a simple explanation of phenomena might be wrong, precisely because it's incomplete. More generally, Occam's Razor has influenced scientists in the direction of simplicity, some say to the exclusion of accuracy. For example, it may just be impossible to provide a systematic, simple account of all natural laws – a so-called theory of everything.
To review, see The Scientific Method Explained by a Scientist.
What criteria do scientists use to choose among competing hypotheses?
There are several criteria for determining the superiority of one scientific theory over another. These include observational consistency, a theory's predictive power, how well the theory explains the relevant underlying causal mechanism(s), the theory's fruitfulness in making surprising or unexpected predictions, and the theory's simplicity and coherence.
To review, see What Makes One Scientific Theory Better Than Another?.
Causal reasoning is fundamental to empirical knowledge. To connect events in such a way as to explain nature, be it in terms of so-called natural laws, such as gravity, or in terms of biological theories, such as evolution, is to enlist the concept of causation. If you know the causes of objects and events, you are better able to determine how or why something occurred. This knowledge facilitates asserting control. If you know the causes of events and objects, you are better able to manipulate the outcome you want. It is a method for making predictions. If you know the causes of events and objects, you are better to foresee their occurrence or behavior in the future, as opposed to merely guessing.
Not only does the concept of causation explain why things happen, it can also be enlisted to determine responsibility. If you know the causes of objects and events, you are better able to determine who is to be praised or blamed for some occurrence. There is a scene from the 2004 film, "Collateral" that highlights the relationship between moral responsibility and causal explanation. In the scene, a taxi driver played by Jamie Foxx has just dropped his fare off (Tom Cruise's character) outside an apartment building. Moments later, a body falls on top of the taxi. As Foxx's character scrambles out of his taxi to check on the fallen person, Cruise's character walks around the corner. In a flash, Foxx's character makes the connection. "You killed him", he says in disbelief. "No, I shot him", Cruise's character responds. "The bullets and the fall killed him". Here, Cruise's character attempts to shift his responsibility for killing a man with the proximate cause of the man's death – the bullets and the fall.
So, just what is a cause? Another way to ask the question is, what is a cause and effect relation? A cause is said to be that which, when it obtains, gives rise to another event. Cause and effect relations may be perceptibly discrete, such as a knife and a cut on the skin. Others are not perceptibly distinguishable, such as heat and fire. What makes one set of events a causal relation, and another set a mere coincidence, is part of what we consider knowledge. Indeed, causation is the foundation of the theoretical and experimental sciences, which contribute to our repository of knowledge. Theoretical science is the practice of proposing theories about how nature works. The conceptual framework for how we think about things occurs in this activity. Einstein's theorizing provided experimental science with ideas to test. Creating good experiments, of course, significantly impacts the conclusions drawn. Because causal relations "in real life" are wildly complicated – there are typically too many variables for which a researcher can account – controlled experiments, while artificial, allow us to more clearly see causal relationships. Controlled experiments typically involve control groups, i.e., setups that are identical to the ones in the experiment, except they are not exposed to the factor being tested for the causal relation.
Scientific reasoning proceeds by way of cause and effect analysis. A hypothesis is tested through carefully constructed experiments to see if a causal relation obtains between objects or events, be they proximate or ultimate. A legitimate test of a causal hypothesis is that the prediction must be verifiable. The prediction must not be trivial; and the prediction must have a logical connection to the hypothesis.
Moreover, experimentation is typically controlled: There are multiple experimental setups that differ only by one variable. Consider the current CERN particle accelerator project, which tests different particle physics theories. The Large Hadron Collider, which sits beneath the ground in a tunnel covering a circumference of 17 miles, was built specifically to carry out experiments that attempt to answer, among other things, causal questions about particles. It is one example – albeit a physically enormous one – of experimental science.
Requirements for causal relations are:
Let's revisit the concepts of sufficiency and necessity. It is arguably the most common language used in discussions about cause and effect relations, be they natural or artificial:
A related way of talking about causation is an inference to the best explanation. When researchers conclude that a hypothesis is the best explanation for a set of facts, they at least sometimes mean that it is the most probable cause. Not all explanations are equivalent, however, to causal inferences. A doctor who says, "I won't prescribe antibiotics, because you have a viral, not a bacterial, infection", offers an explanation for why you will not receive a prescription for an antibiotic. (Here, the point is not to say that the explanation is a case of an inference to the best explanation, but rather to distinguish explanation from causation).
On the other hand, a doctor who says, "Your throat is sore and your tonsils are swollen because you have strep throat" makes a causal claim – in this case, presumably, the best explanation of the symptoms is the causal claim of strep throat. An example of an inference to the best explanation, which is not a causal claim, could be something like this: "I must have some sort of infection. I've been in bed for over a week and my symptoms are consistent with infection – sore throat, clogged ears, and a headache, all of which were preceded by a fever". Here, the symptoms are taken as evidence best explained by an infection.
A look at 19th-century philosopher, J.S. Mill's methods for determining causal relations can expand our understanding of causal reasoning. Mill formulated five experimental tests of causal relations:
The method of difference involves looking at a situation in which the relevant elements are found to be identical in all aspects identified, except one case. In that one case, an event occurs – an effect – that has not occurred in the others. The difference between the other identified aspects of the situation and the one in question reveals the cause of the effect. Consider the following situation:
The method of difference does not guarantee the identification of a cause. The problem with the method of difference is there can be numerous differences that could account for the effect. Ursula could have become ill from something else, for example. Part of the issue is determining, in this case, at least, the relevant time slice to consider.
The method of agreement looks, as the name suggests, for a commonality, not a difference. If, in all cases where an effect occurs, there is a single prior factor, X, that is common to all those cases, then X is the cause of the effect. Consider the following situation:
Here again, the method does not provide a guarantee that the cause has been identified. It does not account for other possible causes, circumstances that, though common to the people who became ill, were not considered as potential causes. These could include, for example, having drunk the same water, or being exposed to a bug, etc.
What we can see so far is that the methods highlight the desirability of a controlled experiment. "Real life" situations are remarkably complex. There are many variables unaccounted for, so a controlled environment, combined with the right sort of hypothesis and experiment, increases the likelihood that the correct cause and effect relation can be identified.
By combining the first two methods together, the joint method of agreement and difference provides a more robust test of a causal relationship. This method involves comparing situations in which the commonalities and differences are sifted out. Mill explains the method this way: "If two or more instances in which the phenomenon occurs have only one circumstance in common, while two or more instances in which it does not occur have nothing in common save the absence of that circumstance, the circumstance in which alone the two sets of instances differ, is the effect, or the cause, or an indispensable part of the cause, of the phenomenon". Here is a situation in which the joint method is employed:
The common element among Hector's horses is they are all fed grain. They also drink out of water buckets that are filled multiple times a day. The only difference between them is that Champ is fed a grain with a lower protein percentage than his stable mates. One morning, Champ does not eat his grain, but Buster does not eat his, either. The percentage of protein should not be the issue, and with all other variables being accounted for, Hector notices that both Champ and Buster's water buckets are empty. The cause of their refusal to eat their morning grain is determined to be dehydration. For this causal relation to be strengthened, Hector would need to conduct experiments (which he likely would not want to do, in order to protect his horses' wellbeing and happiness), in which water is withheld for a certain period before feeding time. Moreover, it should not be the case that the horses refuse their morning grain when they are fully hydrated. In short, the joint method of agreement and difference requires that the effect obtains whenever the cause is present, and that the effect is absent when the cause is absent. Another way to put the relation is as follows: the cause is both necessary and sufficient for the effect, as the cause guarantees the effect, and without the cause, the effect does not obtain.
The method of residues is, Mill declares, "a peculiar modification of the method of difference" (A System of Logic, p. 490). It relies on previous knowledge of certain causes and effects. It also involves a more complicated notion of a cause – it assumes more of a causal cluster for a given effect. To take a simple example, suppose you have a cat, Maurice, who is incredibly nervous about being away from home, and even more upset about going to the dreaded veterinarian's office. So terrified is Maurice that you cannot let him out of his carrying case in order to weigh him. Worse yet, suppose that he howls and yowls in agony whenever you put the carrier down. The vet needs to weigh Maurice, and asks you to hold the cat in the carrier while you step on the scale. You already know how much you weigh, and you already know how much the carrier weighs – you were prepared for this drama, given Maurice's consistent behavior over years of visits to the vet. So, your weight, and the carrier's weight is subtracted from what the scale reveals – Maurice's weight is the residual number. Here is a situation in which the method of residues is employed:
The method of concomitant variation is employed when we want to understand a causal relation in terms of proportionality. We take for granted, for example, that an intense headache will be alleviated by a painkiller, high cholesterol lowered by statins, and a reduction in salt intake generally correlates with a reduction in blood pressure. A variation in a cause, in this method, sees a concomitant variation in effect. The relation can be inverse or parallel. Here is a situation in which an anticoagulation nurse – a nurse who specializes in anticoagulant therapy – uses the method of concomitant variations to determine the dose that will yield a blood test that comes back between the values of 2.0 and 3.0:
To review, see:
What is the difference between correlation and causation?
Causal relations include correlations, but correlation is not the same as causation. A correlation between events is temporal, i.e., events tend to occur at or around the same time. These events might also be related in some way. So, for example, there is a causal relation between the speed a car travels and gasoline consumption while there is only a correlation – in this case, a temporal connection – between a rooster crowing and sunrise. After all, a rooster doesn't cause the sunrise and, even though roosters may begin crowing at sunrise, they also crow throughout the day. So, it would be erroneous to say the sunrise causes a rooster to crow.
To review, see Correlation and Causation.
What are some ways causal relations are visually represented?
There are several ways causal relations can be diagrammatically represented.
Causal network diagrams use arrows to show how events are causally related:
Fishbone diagrams use a horizontal presentation to show how causal factors contribute to an effect:
To review, see Causal Diagrams.
What are some fallacies associated with causal reasoning?
False cause is the weak counterpart to cause and effect arguments. In this fallacy, a causal relationship is asserted between premises and conclusion, when it most likely does not exist. One way to detect this mistake is to ask whether or not a more plausible cause could be found for the supposed effect. Below are some examples of false cause.
Example 1:
Every time I go to see a movie, the lights go down. Therefore, I cause the lights to go down at the movie theater.
Example 2:
Every morning when I get up, my cats go running down the hall to the kitchen and sit by their food bowls. My waking up must cause them to run down the hall.
Example 3:
I kissed a frog about a week ago, and now I have a wart on my hand. You shouldn't kiss frogs because then you'll get warts!
Just because something happens every time I do something does not mean that I am the cause of it. A better explanation of why the lights go down in the movie theater is that someone who works in the theater dims them, because it's time for the movie to begin.
Similarly, my waking up may be a sign to my cats that breakfast is about to come, but the more probable cause of their running to the kitchen is their expectation of being fed. In fact, most cats run to the kitchen whenever someone heads in that direction, regardless of whether or not that person's just awakened!
Finally, there is no scientific evidence to support the myth that kissing frogs will give you warts. Indeed, the myth is about toads, anyway, and even then there is no evidence to conclude that kissing a toad will cause warts. A dermatologist can provide a much more plausible cause.
To review, see Correlation and Causation.
This vocabulary list includes the terms listed above that you will need to know to successfully complete the final exam.