The Process of Science

"What Is Hypothesis-based Science?"

Generate Hypotheses
If you are trying to answer the question of how to prevent colds, most of the processes shown here will be at work.  You will observe how others try to avoid colds and if they were successful.  You will draw on you own past experience with colds and perhaps think of scientific studies on the subject.  Then with some imagination, intuition and perhaps a bit of luck, your brain will come up with an idea of what might decrease your chance of getting a cold.  If you have been a careful observer and acquired some valid information about colds, your idea will be a reasonable hypothesis of what prevents a cold.

Testable Hypotheses
To meet the standards of science, a hypothesis must be testable.  Some potential hypotheses clearly cannot be tested.  For example, I might propose that my colds are the result of someone practicing voodoo on me.  This would be a supernatural explanation based on forces that are not part of the material world.  Science can only measure that which is part of nature, so supernatural phenomena could never be studied or predicted by scientific methodology. 

On the other hand, I might propose that I get colds due to becoming chilled after swimming.  This may not be true, but it is a hypothesis that can be tested.

Falsifiable Hypotheses
A hypothesis must also be falsifiable.  That is, there must be a possible negative answer.  For example, if I hypothesize that all green apples are sour, tasting one that is sweet will falsify the hypothesis.  Note, however, that it is never possible to prove that a hypothesis is absolutely true.  If I set out to prove that all green apples are sour, how many would I need to taste?  Even if I tasted thousands of them, it is always possible that the next apple would be sweet.  Science does not give absolute answers to questions, but it is possible to validate hypotheses as true beyond a reasonable doubt.  So if all thousand green apples that I tasted were sour, I would assume that my hypothesis was, for all practical purposes, true.

What is an example of a hypothesis that is not falsifiable?  Questions dealing with ethics, morals, or justice fall into this category.  I could hypothesize that cheating on an exam is wrong, but this is a question of ethics, not science.  Questions if this type are not falsifiable and should be answered by philosophy or religion.

When formulating a good hypothesis, many elements should be considered.  First it is crucial to base a hypothesis on careful observations.  Faulty input when constructing an answer to a question will almost certainly lead to an incorrect answer.  It is also important to define the problem clearly.  If you don’t know exactly what the question is, you are not likely to find an answer.  Scientists try not to jump to conclusions.  No matter how obvious the answer to a question might seem, it is still necessary to formulate and test a hypothesis.  There is a tendency for hypotheses to become more complex than necessary.  When seeking an answer to a question, the simplest solution is most likely to be the correct one, so simple hypotheses should be considered first.

Formulation of hypotheses is based on logic, but be forewarned that what seems logical is not necessarily true.  For an extreme example, consider Aristotle’s hypothesis that the sun and planets circle around the earth.  This idea seemed logical at the time, since it explained why the sun rose in the east and set in the west each day.  It took 2000 years and the development of the telescope for this logical hypothesis to be disproved.

Inductive Reasoning
Let’s say that you decide to construct a hypothesis on how to avoid colds.  Your dietician has told you that fruits and vegetables contain vitamin C and you have observed that your friends that eat many fruits and vegetables get fewer colds.  You have also read an article in a medical journal that describes how vitamin C reduces throat and nose irritation.  So you formulate the hypothesis that consuming vitamin C decreases the risk of catching a cold.  To arrive at the hypothesis you have used inductive reasoning.  That is, you have combined a series of specific observations to discern a general principle.  Your hypothesis is a good one, since it testable, falsifiable, and based on logic.  However, you will not know whether it is true or false until you test it.

"How Are Hypotheses Tested?"

Experiments have been performed by scientists for several hundred years.  This wood carving from the 18th century shows early biologists attempting to determine whether exposure to electricity affects the growth of plants or the activity of animals.  Experiments are considered to be the most rigorous way to test a specific hypothesis.  The experimental method is usually preferred because it allows the scientist to control conditions under which a given phenomena takes place.  Manipulation of the environment of an experiment provides a way to minimize the number of alternate explanations for the data and increases the likelihood of arriving at the correct conclusion.

The term “control” has a very specific meaning to scientists.  In the experimental method of hypothesis testing, the experiment is divided into two parts.  In one part, the subject of the experiment receives a treatment designed to elicit a response related to the hypothesis.  The experimental control is handled identically, except that the treatment is not given.  If the treated and control groups differ, then the difference is probably due to the experimental treatment. 

In the example shown here, the scientist is trying to determine whether echinacea tea is an effective treatment for colds.  The hypothesis is this:  “Drinking echinacea tea will relieve cold symptoms if taken during the early stages of a cold.”  To set up an experiment to test this hypothesis, the scientist locates two groups of people who have come to a clinic for treatment of early cold symptoms.  One group is given echinacea tea to drink for the next few days, whereas the other group (the control) is given colored water that only looks like tea.  This is an attempt to make the two groups as similar as possible, so that the only difference in their attempts to get rid of the cold will be the consumption of echinacea.  The sham tea is called a placebo.  That is it looks the same as the experimental tea and even has the same water content, but should have no effect on colds.  People in the two groups did not know if they were given echinacea tea or the placebo.  This is important because just knowing that they are being medicated often causes patients feel better.  So the scientist designing this experiment hopes that the only independent variable relative to cold severity is the echinacea compound.

In this experiment, the patients that drank the real tea did report a bit more relief from cold symptoms than those in the control group.  So the scientist concluded that the tea was somewhat effective at relieving cold symptoms.  However, one experiment is seldom enough to convince the scientific community that an effect is real.  In the case of Echinacea tea, additional experiments have cast doubt on these results and it is still not known for sure whether drinking the tea is an effective cold treatment.  You will learn more about experimental controls and variables in this week’s reading assignment.

A scientist using the experimental method must be careful that bias does not creep into the process.  This can happen if the person performing the experiment has an opinion on what the result should be and unconsciously pushes the results in that direction.  To avoid this, the experimenter is often “blind” to the hypothesis being tested and to which group is the experimental one.  In the case of echinacea tea, the best experimental design would utilize a technician with limited knowledge of the hypothesis to conduct the experiment.  A “blind” technician would not know which persons received the real vs. sham tea and thus would not be biased when giving instructions to the subjects or when interviewing them later about the severity of their cold symptoms.  In a so-called “double-blind” experiment, neither the technician nor the subject know which tea the subject has received.  If the subjects know little about the hypothesis or the nature of the tea, they are more likely to report the results of their treatment accurately.

Alternate Hypotheses
Now think back to the hypothesis that people who take vitamin C will have fewer colds than those who do not take the supplement.  Let’s say that this prediction was tested by an experiment in which two groups of people were given supplements to take for 3 months with one group receiving vitamin C and the other group a placebo.  If the results showed that the people taking vitamin C suffered fewer colds, we would conclude that the prediction was true.  But before assuming that the hypothesis is correct, additional tests should be conducted to assure that the results are reproducible.  We should also consider alternative explanations for the results.  For example, were the two groups of people equally healthy at the beginning of the experiment?  Did the groups differ in other aspects of their behavior that might have influenced the results?  Perhaps those in the control group worked in places where they were exposed to more cold viruses.  A scientist should not make a firm conclusion about the correctness of a hypothesis until the experimental results have been replicated and most alternate hypotheses have been ruled out.

It is not always possible to use the experimental method to test a hypothesis.  When studying animals in the field or when humans are the subjects, it is difficult to set up an experiment and is often impossible to control the variables.  The correlative method offers a second approach to testing hypotheses. 

A classic example of the correlative approach is the work of John Snow during the London cholera epidemic of 1854.  Snow hypothesized that cholera was spread by contaminated water, but he had no way to prove it.  So he obtained a map of London and placed a dot at the location of each cholera victim.  This eventually showed a spatial correlation between the incidence of cholera and a certain water source.  Click the button to read more about this important correlative study.

Statistics are widely used by researchers in both the social and natural sciences.  In the experimental method, statistical tests allow the scientist to evaluate whether the results of a single experiment demonstrate an effect of the treatment.

Significance has a special meaning to scientists.  It does not mean important, as in “Graduation was a significant event in my life.”  Instead it is used in conjunction with statistics to indicate the probability that two groups differ by chance.  If the probability is very low, then than the groups are said to be significantly different from one another.  In the experimental method, a significant difference between control and experimental groups means that the experimental treatment had the predicted effect.

Statistical Significance
Here are the results of an experiment in which the effect of zinc lozenges on duration of colds was tested.  Researchers at the Cleveland Clinic performed the study on a group of employees.  Each subject began the experiment within 24 hours of developing cold symptoms.  Half the cold sufferers were given zinc lozenges and half placebos in a double-blind design.  As you can see from the bar graph, the average recovery time for the control group was 7.5 days as compared to 4.5 days for the experimental group.  It appeared that zinc shortened the average duration of colds by 3 days.  However, without the use of statistics, this conclusion cannot be considered valid.  For example, what if only 2 employees were in the experiment?  The fact one person recovered 3 days sooner that the other could certainly be due to chance rather than to zinc consumption.  If 20 employees had been used, we would be more impressed with these results, but what if the decrease in cold duration was only 1 day or 0.5 day?  When should we conclude that zinc truly has an effect?  Statistical tests can give us an answer.  In this case, 100 employees took part in the test, and statistical analysis revealed that the probability of the difference being chance was only 1 in 10,000 (0.01%).  Thus, the conclusion that zinc speeds recovery from a cold seems to be valid.  But note that statistics cannot judge the quality of the experiment.  If an experimental design is poor or if it is performed incorrectly, statistical tests cannot evaluate the problem and might even indicate statistical significance.  So never take statistical results and the “final word” on an experimentally tested hypothesis.  

"How Is Scientific Information Evaluated?"

Now that you have learned how difficult it is to establish a scientific “truth”, we will examine the publishing procedure.  It takes quite a long time for a scientist to progress from the first experiment to publication of a conclusion, since rigorous controls must be used to rule out alternative hypotheses, samples must be large enough to avoid sampling error, and tests must show statistical significance.  When a scientist is finally able to draw a conclusion, a paper is written and submitted to a scientific journal.  There are hundreds of different journals in the field of biology alone, and the author must pick the one best suited for the specific area of research described in the paper.  When the journal receives the paper, it does not just assume that the conclusions are correct.  Instead, the paper is sent to other scientists working on the same problem for review.  The analysis and criticism of one scientist’s work by others is called peer review.  The paper with the reviewer’s comments attached is then returned to the author.  If reviewers think that the experimental design of the research was poor or if the subject is not appropriate for the journal, the paper may be rejected.  More often, the author is asked to make revisions in the paper and resubmit it.  If the author satisfactorily addresses the peer reviewer’s concerns, the paper will then be published.  These journal articles, describing the original scientific work, are called primary sources. In future college courses, you may be asked to write a term paper and use primary sources for some of your references.  So remember what constitutes a primary source in the sciences.

Secondary Sources
When an especially relevant or interesting article appears in a journal, it may be reported in the media.  Such reports can appear in newspapers or magazines and are often heard in radio or television newscasts.  The Internet has also become a source of news reports, with comments on scientific findings appearing in legitimate news stories as well as in politically motivated websites and blogs.  All of these media reports are considered secondary sources of information.  They merely summarize or comment on the findings of others and do not represent the original work.  Because these sources are designed for viewing by the general public, they usually represent  a simplified version of the original work and use language that non-scientists can understand.  Thus, secondary sources of information form most people’s opinions about scientific results.  When you write papers on scientific topics in future courses, you no doubt will use some secondary sources, but be sure to scrutinize them carefully to assure that they are reliable.