The following post is an extension of a much earlier post of mine called Teaching the Hypothesis. If you haven’t read it yet, you might start there. Here I focus on the distinction between generalizing and explanatory hypotheses and how we use them in the science practices.
In 1991 I had moved to Seattle to begin my high school science teaching career and was also training as a competitive distance runner with the local post-collegiate and semi-pro runners in the area. Over meals out and meals made together after long, hard runs I quickly learned from the others about the recently reported performance benefits of a diet high in omega-3 fatty acids. Being in the Pacific Northwest it was easy to load up on excellent sources of omega-3s like salmon and halibut, but for a hopeful additional edge against the competition, many of us also took daily fish oil supplements.
Even though I had majored in biochemistry in college and had a secondary science teaching certification, I was unfortunately not trained as a scientist, let alone to think like one. So if my running buddies, especially those far faster than me, claimed that fish oil would make me faster, then it was true.
What the endurance athletics community was all a buzz about was a provocative 1988 study published by two medical scientists at the University of California at Irvine, David Leaf and C. R. Rauch. Leaf and Rauch had investigated how nutrition affected aerobic performance. One result they described in their paper was a pattern they discovered: increasing dietary omega-3 fatty acids seemed to cause an increase in aerobic performance in athletes. However, Leaf and Rauch had not proven that the pattern was true, they had simply shown experimental support for the pattern. But as a young science teacher, I had no idea how to carefully communicate this claim to my biology students and the runners I coached on my school’s cross country team. During a nutrition unit or when discussing with my athletes what to eat as a runner I likely said something like, “And did you know that eating foods high in omega-3 fatty acids like fish will make you faster?”
I’m better now. I continue to learn from my mistakes as a science teacher (you can read more about me as an impostor here), but I also now have more experience doing science myself as well as more experience teaching it. I now use this story, and a study that was published nine years later, as one of many ways to focus my students on the Nature of Science and how science works. But more specifically, my goal is to provide an illustration for my students from the literature on the difference between a generalizing hypothesis and an explanatory hypothesis, and between predictions and experimental hypotheses in general.
Generalizing and Explanatory Hypotheses
It is important to make clear here that there are two distinct categories of experimental hypotheses. First, there are generalizing hypotheses that are descriptions of observed patterns in nature. For example, dietary omega-3 fatty acids increase aerobic performance. A pattern may be real and meaningful and repeatable across broad natural landscapes. A pattern may also be real, but not meaningful. And a pattern may be accidental, where one variable may be associated with another variable but neither causes the other (i.e. correlation does not imply causation). Then there are explanatory hypotheses, possible explanations for observations in nature, that explain why the potentially meaningful patterns are there or how the patterns are generated and maintained.
Results must be Repeatable to Maintain Credibility
Three Scandinavian exercise physiologists were so intrigued by the pattern discovered and reported by Leaf and Rauch that they set out to replicate the Leaf and Rauch study and test their hypothesis, but under more controlled conditions. The scientists, Raastad, Hostmark, and Stromme, published their work in 1997 in the Scandinavian Journal of Medicine and Science in Sport (Figure 1). The link to the article is here. It is behind a paywall, but perhaps your institution has a way around it. You could also email Truls Raastad directly for a copy.
What I find so interesting and instructive is how Raastad et al. moved so elegantly and logically through scientific methodology and I share this with my students. In fact, I have my students read the paper while following the guide in Figure 2 below.
To help my students dial in on the hypothesis component of the Nature of Science, I focus them in on a few areas of the paper, as illustrated by #3 and #4 in Figure 2 above: What are the generalizing and explanatory hypotheses mentioned in the Raastad et al. paper, and which hypothesis is actually being tested?
The Hypothesis on Trial
In Figure 3 below is the first sentence (and first paragraph) of the Raastad et al. Introduction. Notice how they get right to the pattern claimed by Leaf and Rauch. They also make the statement tentative: “may.” However, words like “may” aren’t required of hypothesis statements. Indeed, being tentative is implied by the fact that a hypothesis statement is falsifiable.
At the end of the Introduction, Raastad et al. then propose a mechanism (explanatory hypothesis) to explain the Leaf and Rauch pattern, if indeed it is real, and restate the generalizing hypothesis (from Leaf and Rauch) that they themselves are retesting (Figure 4).
Clearly, Raastad et al. propose that a possible physiological explanation (an explanatory hypothesis) for the suspected pattern is that
“[O]mega-3 fatty acid supplementation may improve performance by increasing tissue oxygenation due to reduced blood viscosity, thereby increasing maximal cardiac output and the peripheral blood flow.”
The three exercise physiologists probably had no intention initially of testing their explanatory hypothesis because of its complexity and how invasive it would be for their test athletes. However, if they had found strong confirmation for the Leaf and Rauch pattern, it may have given them a reason to further pursue the underlying mechanism for the pattern; i.e. the explanatory hypothesis.
I use the Raastad et al. article to teach my students much more about the Nature of Science (see again Figure 2), but what I emphasize at this point for my students is the difference between generalizing and explanatory hypotheses, and that for the results of a study to be credible, other researchers must be able to replicate the study and get the same or very similar results.
Relating Generalizing and Explanatory Hypotheses
Using an example familiar to most folks who have taken a biology course, Figure 5 illustrates how generalizing and explanatory hypotheses are related.
In the 1850s, Hungarian monk, Gregor Mendel, formally described some inheritance patterns he and others had noticed: Forms of traits become segregated from each other during sex cell formation and different traits sort independently. Then Mendel tested if the patterns were real. And he tested them and tested them and tested them again. In the end, after the patterns continued to exist, Mendel proposed the Laws of Segregation and Independent Assortment. Alongside the patterns, he proposed a reason for the patterns and hypothesized that the hereditary material was not fluid, but instead particulate in nature. The particulate explanation allowed for the prediction that the traits Mendel was following should not behave like a fluid, but should stay distinct and recombine in later generations—and indeed, in predictable proportions. This hypothesis was further tested by T. H. Morgan who lifted Mendel’s particulate hypothesis to a more sophisticated level and, with his results of his X-linked white-eye mutation experiments, Morgan developed the Chromosomal Theory of Inheritance.
Figure 5 thus shows how sometimes, but rarely, explanatory hypotheses can mature to the level of scientific theories, and how generalizing hypotheses can sometimes prove universal enough to be called laws. But while theories can sometimes explain why laws work the way they do, they themselves can never become laws. Theories aren’t patterns, they are explanatory frameworks for patterns. A similar version of Figure 5 appears in my 2015 American Biology Teacher paper, Hypothesis Generation in Biology: A Science Teaching Challenge & Potential Solution.
Hypothesis Testing in the Science Classroom
In the Science Classroom, most of what our students do is the testing of patterns, in other words, the testing of generalizing hypotheses. Only when patterns are supported by experimental evidence do we encourage our students to come up with mechanisms to explain the patterns. In some cases, students can test their explanations with additional controlled experiments, but often this can take time and some pretty sophisticated methods and equipment. What is most important, though, is that our students constantly practice their skills at designing and running controlled experiments, analyzing, summarizing, and interpreting data, and coming to logical and arguable conclusions, regardless if the hypotheses under fire are descriptions of patterns or mechanistic explanations.
An Example from Physiology
Let’s take “physical activity increases heart rate” as an example of a testable hypothesis. This statement is a generalizing hypothesis, a pattern that we may observe and take note of in ourselves and in a group of test subjects: “It seems that when this group of people exercises, all of their heart rates increase.” We then formally describe the observation with a generalizing hypothesis.
Hypothesis: Physical activity increases heart rate.
This statement has all the appropriate pieces of a hypothesis: it is tentative (can be found to be false), it is testable, and it is within the boundaries of nature. We can thus do a controlled experiment to test if the pattern is real and repeatable among many subjects from different groups and even across animal species. The particular experimental methods will allow us to make a critical prediction from the hypothesis.
Prediction: Subjects who exercise vigorously for two minutes will have significantly higher heart rates than when they started and will also have significantly higher heart rates than a resting treatment group.
We might even add the hypothesis at the beginning and have what the science education literature calls the Research Hypothesis. Research hypotheses are rarely seen in published experimental studies, but they can function as important organizing tools and take on the form, If X hypothesis is valid, and I perform Y methods, then I can predict Z as a specific, measurable outcome.
Research Hypothesis: If physical activity increases heart rate, and ten subjects exercise vigorously for two minutes while ten other subjects sit quietly, then the subjects that exercised will have significantly higher heart rates than when they started and will also have significantly higher heart rates than the resting treatment group.
If we continue to test the pattern over and over again and in different species, at some point we may even conclude that the positive relationship between exercise and heart rate is simply a biological law and we may even be able to explain the relationship mathematically.
But, what we don’t yet have is a physiological explanation for why the pattern exists or a mechanism for how it is produced. The explanation for why there is a repeatable pattern is our explanatory hypothesis. Explanatory hypotheses are often difficult for students in a lab setting to come up with, let alone test in any meaningful way. However, we can still propose explanatory hypotheses without ever testing them—students could do this in their conclusion and evaluation of their lab paper and scientists do this all the time in their discussion sections.
For example, students in an IB or AP Biology course or a college Anatomy and Physiology course may propose the explanatory hypothesis that increased exercise increases carbon dioxide production and lowers blood pH which in turn stimulates the heart to beat faster, thus removing excess carbon dioxide and bringing blood pH back to normal. This is a testable explanatory hypothesis but we would not expect our students to do the controlled experiment, at least not on their peers, but maybe on mice.
An even Simpler Example from Physical Science
Students are shown three swinging pendulums of the same mass and notice the pattern that the longer the pendulum, the slower the swing.
Hypothesis: There seems to be a positive relationship between the length of a pendulum and the period of its swing.
Prediction: If I increase the length of a pendulum by 0.2 meters at a time, beginning at 0.2 m and ending at 1.0 m, swinging each different pendulum with the same beginning angle, then the number of seconds it takes the pendulum to make one pass through its arc will also increase… and might even be linear.
Notice that the prediction is specific to the planned test/experiment.But also notice that the prediction is not simply a restatement of the hypothesis: There will be a positive relationship between the length of a pendulum and the period of its swing. Instead, the prediction includes a clear, unique, planned test that should produce a specific measurable outcome.
Beginning with a simple experiment like this can lead students to ask additional questions about pendulums and propose additional interesting hypotheses and predictions.
In summary, science tests its hypotheses better than any other human endeavor. Scientists do their best to test (even falsify) their own hypotheses before they will accept them. Helping students understand that there are two types of hypotheses and how to use them in science inquiry can clear up some misunderstandings about hypotheses and predictions and open up a lot of possibilities in the classroom laboratory where students are learning how to think like scientists.
Leaf, D. A., & C. R. Rauch. (1988). Omega-3 supplementation and estimated VO2 max: a double blind randomized controlled trial in athletes. Annals of Sports Medicine, 4: 37-40.
Raastad, T., A. T. Hastmark, & S. B. Strømme. (1997). Omega‐3 fatty acid supplementation does not improve maximal aerobic power, anaerobic threshold and running performance in well‐trained soccer players. Scandinavian Journal of Medicine & Science in Sports, 7: 25-31.
Strode, P. K. (2015). Hypothesis Generation in Biology: A Science Teaching Challenge & Potential Solution. The American Biology Teacher77:500-506.