A Scientist’s Guide to Making Brilliant Hypotheses [Part 1]
A Scientist’s Guide to Making Brilliant Hypotheses -Part 1-
Making hypotheses is one of the most important parts of being a scientist, but have you ever wondered what it really means? Like many people, the concept of “the hypothesis” was introduced to me as a kid when learning about the scientific method. And that’s about as far as it ever went. Even when I was studying to become a scientist at University, I don’t recall ever actually having any formal training on the subject.
Science professors have a hard enough time just trying to teach us the fundamentals of chemistry, physics, etc. to get us ready for the work-place, so I can understand why they would skip over something we have already “learned” in high school. For something so essential to scientific research, making hypotheses should be common sense. But what if it isn’t?
I recently started to ask myself this question. I recalled noticing inconsistencies in the way that I, my colleagues, and project stakeholders thought of hypotheses. The more I thought about it, the more I realized there seemed to be a bit of confusion about not only what makes a good hypothesis, but also about what is or isn’t a hypothesis? In fact, I realized that I wasn’t really sure either.
There were two questions that I felt really needed to be answered:
- What is a hypothesis?…
- What makes a good hypothesis?…
What is a hypothesis?
The general definition of a hypothesis, from the Cambridge Dictionary, is “an idea or explanation for something that is based on known facts but has not yet been proved”. That is pretty much what I remember from high school, but this definition is a bit loose for practical use in research. For example, since the surface of the Moon looks yellowish and is a bit “holey”, I could hypothesize that the surface of the moon is made of cheese. By this definition, this is technically a hypothesis…
With some more research, I realized that in practice what we make are more specifically; scientific hypotheses, statistical hypotheses, and predictions. It is the mixing of these three different things which I think has led to the confusion that I have noticed between my colleagues at work. For, while they are similar, scientific hypotheses, statistical hypotheses, and predictions are actually quite different in subtle but important ways.
Scientific hypotheses are testable and falsifiable conjectures proposed to account for the relationship between two or more variables. It is what we believe will happen in the experiment. For any given scientific question, multiple scientific hypotheses could be formulated. We can then attempt to falsify each hypothesis by generating predictions, collecting data, and performing statistical tests to determine whether the predictions held true.
Famous examples of scientific hypothesis are:
- Einstein’s equivalence principle
- Francis Crick’s DNA makes RNA, which makes protein.
Statistical hypotheses are merely statements about whether or not a pattern, trend, or difference is present in your data. Statistical tests are used to distinguish between the null hypothesis and one or more “alternative hypotheses”.
The “Null Hypothesis” (H0) states that there is no pattern, trend, or difference in groups in the data, e.g. there is no significant difference between the groups or no significant relationship between two variables. The “Alternative Hypothesis” (H1) states that there is a distinct pattern, trend, or difference between groups in the data, e.g. there is a significant difference between the groups, or there is a significant relationship between two variables.
Once we have defined a scientific hypothesis, only then can we can decide what kind of statistical hypothesis is appropriate. This is because the statistical hypothesis is really only a tool for trying to understand the natural phenomena that were are investigating in our experiments.
Examples of statistical hypotheses are:
- Null Hypothesis (H0): There is no significant difference in the stability of peanut butter produced with palm oil and peanut butter made without additives.
- Alternative Hypothesis (H1): There is a significant difference in the stability of peanut butter produced with palm oil and peanut butter made without additives.
Predictions are statements about the patterns, trends, or differences between groups. Although predictions might just seem like another word for a hypotheses, it is in fact quite different. Predictions involve determining the logical consequences of the scientific hypothesis. Predictions, therefore, play a significant role in determine the experimental design.
In order for it to be successful, the prediction should be able to distinguish the scientific hypothesis from likely alternatives; for example, if two hypotheses make the same prediction, observing the prediction to be correct is not evidence for either one over the other. Generally, statistical hypotheses tests are required to determine whether or not a prediction is correct.
For the hypothesis that palm oil improves the stability of peanut butter, an example of a prediction is that in peanut butters made with palm oil less oil should separate from the peanute butter matrix during storage.
Part 2 will look into what makes a good scientific hypothesis.