What is a scientific experiment

In addition to observation and theoretical derivation, experiments are an important part of gaining scientific knowledge. All experiments have similar features and comply with a specific set of rules. Consistency with these rules is important because we need to trust evidence to share it with other scientists. This topic will teach you about scientific experiments and the rules that govern them.

Overview

In the "What is Science?" topic, we discussed what science is and that knowledge can be obtained in different ways — you can derive something theoretically, you can observe some phenomenon, or you can conduct an experiment. An experiment is a method of gaining scientific knowledge in which certain phenomena are studied under certain reproducible conditions. Some differences from observational science are:

1. While observing, the scientist cannot influence the conditions, but during an experiment this is their most important task. To give a simple example, you can watch your cat and find out that she prefers to sharpen her claws before eating. This is an observation. You can also experiment with giving her a few scratching posts and see which one she prefers or changing the feeding times to see if there are differences in behavior. You can also change the food itself or see how other cats behave. So the main difference between a scientific experiment and observation is the ability of a scientist to influence the conditions of the experiment. Sometimes, especially in chemistry, "standard conditions" are used — they differ for different aggregate states, and usually describe a specific pressure, temperature, or concentration.
2. During an experiment, scientists use special devices, such as telescopes, microscopes, test tubes, Petri dishes, etc.
3. Experiments are based on theory and past discoveries. Scientists study the theoretical foundations of the topic they are investigating (by reading scientific papers, for example) and then define the goals and objectives of the experiment. After a series of experiments, a scientist can do a meta analysis, or an examination of data from a number of independent studies of the same subject to determine overall trends.

The general plan of an experiment:

Usually a scientific article is built according to the following plan:

1. Statement of the question, suggestion of possible answers, setting goals and objectives. This is the most important part of preparing for an experiment, because the right question determines the result. The most commonly used is the 0-alternative system: the null hypothesis is that there is no relationship between two observed phenomena. In contrast, an alternative hypothesis suggests that there is a connection. For example, a null hypothesis would be the statement "red cats do not purr louder than other cats," and an alternative hypothesis would be "red cats purr louder than other cats."
2. Creation of a test setup that provides the necessary conditions for to study the phenomenon of interest. Test conditions should test the hypothesis as accurately as possible and not give the opportunity to draw false conclusions. For example, you can choose two groups of cats — red and gray, and as a result of the experiment, you might determine that the red ones purr louder. But in fact, your groups were divided differently — all the red cats were female, and all the gray cats were, and the real result of the experiment could be that female cats purr louder than male cats, and color actually has no affect. So the conditions must be chosen carefully, checking as many facts as possible. All conditions of the experiment must be reproducible — that is, any other scientist must be able to completely reproduce the experiment and get the same results.
3. A direct requirement from the previous paragraph is the creation of a representative sample. First, for the experiment to yield statistically significant results that are clearly not just noise, it is necessary to include a large number of subjects or trials in the experiment. A larger sample size will have greater scatter among the results and will thus more accurately represent the whole population or the full range of outcomes. For example, if you measure the frequency of purring, you might find that one cat has a frequency of 40 Hz. But if we draw our conclusion and report results only based on this cat, we will have incorrect data on purring, because it turns out that the rest of the cats that we didn't study purr at frequencies between 25-30 Hz. Thus, a sufficiently large sample must be examined. At the output, we will get values that differ from each other within a range, and for which one can calculate the deviations. Ideally, the deviation will be small, but if you have a large enough sample, you will be able to find artifacts, or subjects that fall far outside the normal range, like that cat that purrs at such a different frequency, and remove them from consideration. In some experiments, scientists will use a control group — a group of subjects at the same conditions as the experimental group, but not subjected to the effects of the experimental variable. In medical research, one group receives an investigational drug and another group receives an identical dosage form but with no drug — this is called the placebo group.
4. Description of the new phenomenon and its properties, description of the experimental results and their possible interpretation, and statistical analysis. A full discussion of statistics is beyond the scope of an introductory topic, but any scientist needs to have a good understanding of how statistics work and how to apply its laws in practice. In simple terms, the statistical processing of the results allows you to understand if the data samples are similar to each other, which of them stand out too much, and how much you can trust the data. Remember the cat that purrs at a frequency of 40 Hz? We want to remove it from the total sample because it does not fit into the average result, but we also need to explain why exactly it does not fit (for example, he suffered an otorhinolaryngological disease).

Experiment example

Let us consider an article "How cats purr". This nice article about how cats purr will allow us to dip a little into the basics of a scientific experiment (for a more complete understanding of what a Scientific article is, read the corresponding topic). First, at the very beginning we see the introduction, which presents background information about how the author studied publications on the topic and formed a question to which they could not find an answer. How does purring happen and what are its properties? Observation will not be sufficient to answer this question — after all, the mechanism responsible for purring is hidden from view. A theoretical assumption will not help either — there are several alleged mechanisms. So an experiment might help.

The author then described the methodology. First, 10 cats were included in the experiment — this is the sample or series. Further, the author records the purring of all the cats under the same conditions: the same microphone, the same distance to the throat, making sure that all the cats are relaxed and lie in similar positions, and so on. He also investigates the behavior of cats' throats and larynxes. Note that all of the studies were carried out over a long period of time, from when the kitten was 12 weeks old to the age of three years!

Conclusion

We have briefly considered how experiment differs from observation and theory: the experiment uses devices and changes conditions in a controlled manner. In an experiment, sampling is important to reduce the possibility of deviation from real values. The results of the experiment are statistically processed to establish reliability. This allows real correlations to be identified.