Last updated on August 31, 2020
“… Science is constantly proved all the time. You see, if we take something like any fiction, any holy book and destroyed it, in a thousand years’ time, that wouldn’t come back just as it was. Whereas if we took every science book, and every fact, and destroyed them all, in a thousand years they’d all be back, because all the same tests would produce the same result.”
– Ricky Dene Gervais
1. Science as an Ongoing Process
The scientific method can generally be described as the totality of techniques used to investigate phenomena, to acquire new knowledge, or to correct previous knowledge. The investigation methods used are usually based on empirical or measurable evidence. As a whole, science is an ongoing process that usually begins with observations in the environment.
The human being is curious by nature. Everyone who has raised a child or who has dealt with children aged 3-4 years knows it: “Why does the sun exist? Why does it set? Why do the stars shine at night? … Why… Why… Why…” At least during childhood every person has been a little explorer. And it is extremely important to keep up this curiosity because that is what drives science forward.
Also, with adult scientists the cycle begins with asking questions about things they see or hear and then develop ideas (hypotheses) about why things are the way they are. These hypotheses are then tested in a variety of ways. In general, the strongest tests of hypotheses come from carefully controlled and repeated experiments in which empirical data has been collected. Depending on how well the tests match the predictions, the original hypothesis may need to be refined, modified, expanded, or even rejected. If a hypothesis is strongly supported, even a general theory can be developed.
2. Observations in Science
Observation describes the active acquisition of information from a primary source. This can be either through sensory perceptions or through the recording of data with the help of technology. Human senses, however, are mostly exposed to subjectivity as well as errors in perception and are rather of a qualitative nature what makes it difficult to record and compare them. For them it is difficult to meet the three main quality criteria of scientific research:
- objectivity (Is the observation independent of the observer?)
- validity (is the right property measured?)
- reliability (Are the same results obtained when repeating observations?).
The situation is different with measurements. These reduce the observation to a number that can be recorded and compared. Two observations which result in the same number are equal within the resolution of the process. Scientific instruments are designed to measure observations as objectively, as validly, and as reliable as possible.
3. Scientific Question
Scientific questions can be asked in two different ways. In one approach the question tries to explain a certain observation: “Why is the risk of medial knee osteoarthritis increased in patients with varus malalignment?” The other approach is a rather open question: “What are influences of the varus malalignment in the human body?”
However, not only our own observations but also results from previous experiments are included in the question. A precise research on the corresponding problem in specific databases is therefore essential in this phase of the scientific method. If the answer has already been investigated, another question can be based on it. The determination of a good scientific question can sometimes be exceedingly difficult, but it is very important, because the final result of the research will be influenced by it to a large extent.
4. Hypothesis
A hypothesis is an assumption that serves as an answer to the scientific question. However, these assumptions are by no means plucked out of thin air but are based on previous observations and findings as well as logical conclusions. Hypotheses in general can be quite specific, e.g. “The risk of medial knee osteoarthritis in patients with varus malalignment is increased due to a high external knee adduction moment” or they can be broad, e.g. “…due to external loads acting upon the human body”.
It is important for the hypothesis to be formulated in such a way that it can be falsified. This has the theoretical background that in the scientific and statistical sense a hypothesis can never be confirmed. Since in science it is never possible to assess the entire population, it is not excluded that a different result can be obtained with a different sample selection. Therefore, statistical methods always use a so-called null hypothesis (H0) and an alternative hypothesis (H1), which are exact opposites. H0 is the assumption that the scientific hypothesis is wrong (e.g. “The risk of medial knee osteoarthritis in patients with varus malalignment is not increased due to a high external knee adduction moment”). However, from previous studies the scientist suspects that the alternative hypothesis (“… is increased due to a high external knee adduction moment”) is correct and therefore tries to show that H0 is wrong by means of an experiment. If the result of this experiment is significant, H0 is rejected and H1 is accepted for the sample. Strictly speaking, however, this is not a conclusive proof that H1 applies to the entire population.
5. Testable Predictions
Every useful hypothesis makes predictions possible. These could be for example the result of an experiment in a laboratory or the observation of a phenomenon in the environment. Also, the prediction can be of statistical kind and deal only with probabilities.
It is essential that the result of testing such a prediction is currently unknown. If it is already known, the requirements of a scientific hypothesis were not met beforehand and the outcome can be called a consequence. In this case, an investigation would be pointless and should already have been considered while formulating the hypothesis.
If the predictions are not accessible through observation or experiments, the hypothesis is not yet verifiable and thus remains unscientific. A new technology could make the necessary experiments feasible in the future. This would enable testable predictions and would allow the hypothesis to be assigned to science.
6. Gathering of Data
“Without data, you’re just another person with an opinion” – William Edwards Deming
In order to verify the predictions, relevant data must be collected, although the sources of this data can be very different. They can either be taken from the literature (e.g. for reviews), from new observations, or from new experiments. However, thorough testing requires replication to verify the results.
Observation and experiment are fundamentally different from each other. Observations and also measurements in science are targeted recordings of phenomena in the environment. Conclusions can be drawn from these to describe systems and relationships. Likewise, fundamental scientific laws can be determined, but observations cannot proof laws ultimately. The classic stork-baby-example illustrates this: Observations show that the number of new-borns per year is increased in regions where also many storks were seen. As a conclusion one could draw that the storks are responsible for the birth rates. Of course, this is a fallacy, because both increased rates depend on geographical parameters, but this cannot be proven with observations. This so-called cause-and-effect relationship can only be investigated by means of experiments.
As the stork-baby-example reveals, many different factors usually have an influence on the system you want to investigate. These factors are called independent variables and influence the so-called dependent variable, which in the example is the birth rate.
The goal of an experiment is to determine the influence of one of the many independent variables by changing it while keeping the other independent variables as controlled and constant as possible. The change in the dependent variable is then measured and a statement about the influence of the independent variable on the overall system can be obtained.
There are incredibly many types, setups, and designs of experimental studies in science. These differ in their applicability and quality. These will be discussed in a separate article about study designs.
If enough data have been collected in an experiment or observation to be able to make a statement about the hypothesis, it must now be accepted or rejected. In this step it can also happen that the previously formulated hypotheses must be changed, or even extended and new data collections become inaccessible. If a hypothesis could be confirmed many times, this opens the possibility to develop a scientific theory.
7. Development of a Theory
If a certain hypothesis is very well supported by results of many high-quality studies, a scientific theory can be developed to describe a certain aspect of nature. Thereby, simpler theories are preferable to more complex ones because they are easier to test which is a particularly important part of science. Theories must be constantly proven over time, which is why a theory must be falsifiable just like a hypothesis. Even if a theory has been tested and verified several times, it can still happen that results of new scientific studies are not in accordance with it. In this case, a theory must be revised and adapted or even discarded completely. Now, the scientific cycle begins anew.
The most famous example of a disproved scientific theory is probably the geocentric model, which was developed by Aristotle in the 4th century BC and was only replaced by the heliocentrism by Copernicus in the 16th century AD. The heliocentrism was then mathematically defined by Kepler and supported by Galilei’s observations. Today it is a well-recognized theory, tested infinitely often and serves as a basis for research such as space travel and meteorology but also for everyday technology.
