Skip to main content
Biology LibreTexts

1.7: Nonexperimental Scientific Investigations

  • Page ID
    22429
  • Up in Smoke

    You've probably seen this warning label dozens of times. It's been required on cigarette packs in the U.S. since 1965, one year after the U.S. Surgeon General first issued a report linking cigarette smoking with diseases such as lung cancer. The report was based on thousands of research articles, including important research results published by British scientists Richard Doll and Austin Bradford Hill. Starting in 1950, Doll and Hill conducted large-scale, long-term observational studies on smoking and lung cancer and demonstrated a strong correlation between the two.

    cigarette
    Figure \(\PageIndex{1}\) (CC BY-NC 3.0; wikiwand.com)

    Observational Studies

    Many questions in human biology are investigated with observational as opposed to experimental studies. An observational study measures characteristics in a sample but does not attempt to manipulate variables of interest. A simple example of an observational study is a political poll. A sample of adults might be asked how old they are and which of two candidates they favor. The study provides a snapshot in time of potential voters' opinions and how they differ by age of the respondent. Whether the results of the study apply to the population as a whole depends mainly on how large and random the sample is.

    How is an observational study different from an experiment — the gold standard of scientific research studies? The main difference is how subjects are treated. In an observational study, no attempt is made to influence the subjects in any way. In an experiment, in contrast, the researcher applies a treatment to a group of subjects and attempts to isolate the effects of the treatment on an outcome variable by comparing the experimental group with a control group. For example, in 1954, Jonas Salk did an experimental trial of his newly discovered polio vaccine by giving it to a very large sample of children. Children in an equally large control group were given a harmless injection of a saline solution but no vaccine. Salk then compared the two groups of children and determined that the vaccine was 80 to 90 percent effective in preventing polio.

    Types of Observational Studies

    There are three different types of observational studies: cross-sectional, case-control, and cohort studies. All three types have pros and cons.

    1: Cross-sectional Studies

    A cross-sectional study is a type of observational study that collects data from a sample of subjects just once at a certain point in time. The political poll described above is a simple example of a cross-sectional study. A possible link between smoking and lung cancer was also first suggested by cross-sectional studies. Researchers found a higher rate of lung cancer in people who smoked than in those who did not smoke at the time of the study. In other words, the two variables seemed to be associated.

    Cross-sectional studies are relatively cheap and easy to do, but their results are weak, so they are rarely used alone. More often, a researcher uses a cross-sectional study to find variables that may be linked and then does a case-control or cohort study to further investigate a possible relationship between the two variables.

    2: Case-Control Studies

    A case-control study is a type of observational study that compares a group of subjects having a trait of interest (cases) with a group of similar subjects not having the trait (controls). This type of study is retrospective. Subjects are asked to report their behaviors in the past in an attempt to find correlations between specific past behaviors and current status. The retrospective nature of case-control studies is their main weakness. Subjects' responses may be inaccurate because they forget or are dishonest about past habits.

    A classic example of a case-control study is the early research on smoking and lung cancer carried out by Doll and Hill (who is pictured below). In 1950, the two scientists interviewed 700 lung cancer patients (cases) and 700 people without lung cancer (controls). They gathered information on past smoking habits and other characteristics of people in the two groups. When they compared the two groups, they found a strong association between past smoking behavior and current lung cancer status.

    Sir Austin Bradford Hill, portrait
    Figure \(\PageIndex{2}\): Sir Austin Bradford Hill (CC BY 4.0 via Wikimedia.org)

    Austin Bradford Hill was named a British knight for his important research in public health, including his work with Richard Doll establishing a link between tobacco smoking and lung cancer.

    3: Cohort Studies

    A cohort study is an observational study in which a group of similar subjects (the cohort) is selected at the start of the study and then followed over time. This type of study is prospective. The researchers collect data on the cohort periodically for months or even years into the future. Because the researchers collect the information directly, the data are likely to be more accurate than the self-reported recall data in case-control studies. Prospective data also allow researchers to establish the sequence of progression of disease states or other conditions of interest. On the other hand, cohort studies are the most costly and difficult observational studies to undertake.

    One of the largest-ever cohort studies was undertaken by Doll and Hill in 1951. It was based on their earlier case-control study and further investigated the link between smoking and lung cancer. The cohort that began the study included almost 50,000 British male physicians, and they were followed by the researchers over the next 50 years. Initial findings of the study were first reported in 1954, and then updated results were reported periodically after that. The last report was published in 2004, and it reflected on the previous 50 years of research findings. This study provided even stronger evidence for the correlation between smoking and lung cancer.

    Numerous other research studies, including experimental studies, have shown conclusively that smoking causes lung cancer, among many other health problems. The diagram below shows some of the ill effects that have since been demonstrated to be caused by smoking.

    Adverse effects of tobacco smoking
    Figure \(\PageIndex{3}\): Lung cancer is just one of many adverse effects on the human body that research studies have shown to be caused by smoking. The more common adverse effects are in bold. (Mikael Häggström; CC BY-NC 3.0; wikimedia.org)

    Correlation vs. Causation in Observational Studies

    Observational studies can generally establish correlation but not necessarily causation. Correlation is an association between two variables in which a change in one variable is associated with a change in the other variable. Correlation may be strong or weak. It can also be positive or negative.

    • If two variables are shown to have a positive correlation, both variables change in the same direction. For example, an observational study might find that more smoking is correlated with a higher risk of lung cancer. In other words, as smoking goes up, so does lung cancer.
    • If two variables are shown to have a negative correlation, they change in opposite directions. For example, an observational study might find that people who exercise more are less likely to develop lung cancer. In other words, as exercise increases, lung cancer decreases.

    One of the main differences between observational studies and experiments is the issue of correlation vs. causation. Because observational studies do not control all variables, any correlations they show between variables cannot be interpreted as one variable causing another. In experiments, in contrast, all possible variables are controlled, making it safer to conclude that changes in one variable cause changes in another. Unfortunately, when observational studies are reported in the news media, this distinction is not often made. Instead, a variable that is correlated with another in an observational study may be reported incorrectly as causing changes in the other variable.

    In observational studies, it is always possible that some other variable affects both of the variables of interest and explains the correlation. An example of the confusion of correlation and causation in observational studies is the case of the health effects of coffee. Many early observational studies of coffee consumption and health found a positive correlation between drinking coffee and health problems such as heart disease and cancer. Does this mean that drinking coffee causes these health problems? Not necessarily, although news media have reported this conclusion. Looking more deeply into the issue reveals that coffee drinking is also associated with a less health-conscious lifestyle. People who drink coffee tend to practice other behaviors that may negatively impact their health, such as smoking cigarettes or drinking alcohol. Larger observational studies in which such lifestyle differences were taken into account have found no correlation between coffee consumption and health problems. In fact, they have found that moderate coffee consumption may actually have some health benefits.

    Rationale for Observational Studies

    If observational studies cannot establish causation, why are they done? Why aren't all research questions investigated experimentally? There are several important reasons to do observational studies:

    • An observational study may be the only type of study that is feasible for certain research questions because experiments are impossible, impractical, or unethical to undertake. For example, it would be unethical to do an experiment on smoking and health in which subjects in the smoking sample are deliberately exposed to tobacco smoke and then observed to see if they develop lung cancer.
    • An observational study is generally cheaper and easier to conduct than an experimental study.
    • An observational study usually can study more subjects and obtain a larger set of data than an experimental study.

    Models

    Another way to gain scientific knowledge without experimentation is with modeling. A model is a representation of part of the real world. Did you ever build a model car or airplane? Scientific models are something like that. They represent the real world but are simpler. This is one reason that models are especially useful for investigating complex systems. By studying a much simpler model, it is easier to learn how the real system works.

    As a hypothesis, a model must be evaluated. It is assessed by criteria such as how well it represents the real world, what limitations it has, and how useful it is. The usefulness of a model depends on how well its predictions match observations of the real world. Keep in mind that even when a model's predictions match real-world observations, it doesn't prove that the model is correct or that it is the only model that works.

    Modeling Biological Systems

    Many phenomena in biology occur as part of a complex system, whether the system is a cell, a human organ such as the brain, or an entire ecosystem. Models of biological systems can range from simple two-dimensional diagrams to complex computer simulations. The model animal cell depicted in the diagram below is an example of a simple model of a biological system.

    Animal cell structure
    Figure \(\PageIndex{4}\): Model of an animal cell (By LadyofHats (Mariana Ruiz), via Wikimedia Commons; CC BY-NC 3.0)

    A much more complex model of a biological system is the human brain model being developed by the Blue Brain Project. This multi-year, multi-national project is attempting to create a biologically detailed digital model and simulation of the human brain by reverse-engineering the brain down to the molecular level. The aim of the project is to study the brain's architecture and functional principles. Proponents of the model hope that it will eventually shed light on the nature of human consciousness.

    Model Organisms

    Using other organisms as models of the human body is another way models are used in human biology research. A model organism is a nonhuman species that is extensively studied to understand particular biological phenomena. The expectation is that discoveries made in the model organism will provide insights into the workings of the human organism. In researching human diseases, for example, model organisms allow for better understanding of the disease process without the added risk of harming actual human beings. The model species chosen should react to the disease or its treatment in a way that resembles human physiology. Although biological activity in a model organism does not ensure the same effect in humans, many drugs, treatments, and cures for human diseases are developed in part with the guidance of model organisms.

    Model organisms that have been used in human biology research range from bacteria such as E. coli to nonhuman primates such as chimpanzees. The mouse Mus musculus, pictured below, is a commonly used model organism in human medical research. For example, it has been widely used to study diet-induced obesity and related health problems. In fact, the mouse model of diet-induced obesity has become one of the most important tools for understanding the interplay of high-fat Western diets and the development of obesity.

    House mouse
    Figure \(\PageIndex{5}\): The mouse Mus musculus is commonly used as a model organism in human biology research. (Ilmari Karonen, via Wikimedia Commons; CC BY-NC 3.0)

    Feature: Reliable Sources

    You may get most of your news from the Internet. You probably also research personal questions and term paper topics online. Unlike information in newspapers and most television news broadcasts, information on the Internet is not regulated for quality or accuracy. Almost anybody can publish almost anything they wish on the web. The responsibility is on the user to evaluate Internet resources. How do you know if the resources you find online are reliable? The questions below will help you assess their reliability.

    1. How did you find the web page? If you just "googled" a topic or question, the search results may or may not be reliable. More likely to be trustworthy are web pages recommended by a faculty member, cited in an academic source, or linked with a reputable website.
    2. What is the website's domain? If its URL includes .edu, it is affiliated with a college or university. If it includes .gov, it is affiliated with the federal government, and if it includes .org it is affiliated with a nonprofit organization. Such websites are generally more trustworthy sources of information than .com websites, which are commercial or business websites.
    3. Who is the author of the web page? Is the author affiliated with a recognized organization or institution? Are the author's credentials listed, and are they relevant to the information on the page? Is current contact information for the author provided?
    4. Is the information trustworthy? Are sources cited for facts and figures? Is a bibliography provided? Does there seem to be a particular bias or point of view presented, or does the information seem fair and balanced? Does the page contain advertising that might impact the content of information that is included?
    5. Is the information current? When was the page created and last updated? Are the links on the page current and functional?

    Put this advice into practice. Go online and find several web pages that provide information on the topic of smoking and lung cancer. Which websites do you think provide the most reliable information? Why?

    Summary

    • Many questions in human biology are investigated with observational as opposed to experimental studies. An observational study measures characteristics in a sample but does not attempt to manipulate subjects.
    • There are three types of observational studies: cross-sectional, case-control, and cohort studies. All three types of studies have pros and cons.
    • Observational studies can determine whether variables are correlated but not whether changes in one variable causes changes in another variable. Experiments are generally needed to show causation. Although observational studies do not establish causation, there are several important reasons to do them.
    • Another way to gain scientific knowledge without experimentation is with modeling. A model is a representation of part of the real world. Models of biological systems range from simple two-dimensional diagrams to complex computer simulations.
    • Using other organisms as models of the human body is another way models are used in human biology research.

    Review

    1. Explain why observational studies cannot establish causation. Describe an example to illustrate your explanation.
    2. Compare and contrast the three types of observational studies described above.
    3. Identify three possible reasons for doing an observational study.
    4. Why are models commonly used in human biology research?

    Multiple choice. A ____________ study involves the recall of variables that occurred in the past, while a _____________ study involves the observation of variables from the beginning.

    A. positive correlation; negative correlation

    B. negative correlation; positive correlation

    C. retrospective; prospective

    D. prospective; retrospective

    6. True or False. A positive correlation means there are health benefits to the variable under investigation.

    7. True or False. A cohort is a group of subjects of different ages, weights, genders, and health statuses.

    8. A study is done to investigate whether soda consumption influences the development of diabetes. The subjects are individuals recently diagnosed with diabetes compared to controls who do not have diabetes. All of the respondents are asked how many times a week they drank soda over the last two years. Answer the following questions about this scientific investigation.

    a. What type of observational study is this?

    b. The subjects with diabetes are “matched” to the controls, meaning that the researchers tried to minimize the effect of other variables outside of the variable of interest (i.e. soda consumption). What do you think some of those other variables could be?

    c. Do you think the data about soda consumption will be accurate? Why or why not?

    d. How could you change the study to get more accurate data on whether there is a relationship between soda consumption and diabetes? Explain why your new study would be more accurate.

    9. Do you think that computer simulation models of biological systems can be accurate without observations or experiments on actual living organisms or tissues?

    10. Explain why both observational and experimental investigations are useful in science.

    Explore More

    If you are still unclear about differences between observational and experimental studies, watch this short video for a simple comparison of the two major types of studies.

    Learn more about the Blue Brain Project by watching this TED talk.