Experimentation in Scientific Research
by Anthony Carpi, Ph.D., Anne Egger, Ph.D.
Did you know that experimental design was developed more than a thousand years ago by a Middle Eastern scientist who studied light? All of us use a form of experimental research in our day to day lives when we try to find the spot with the best cell phone reception, to try new cooking recipes, and more. Scientific experiments are built on similar principles.
Manipulating and controlling variables are key aspects that set experimentation apart from other scientific research methods. This module highlights the principles of experimentation through examples from history, including the work of Alhazen in 1000 CE and Louis Pasteur in the 1860s.
Terms you should knowtoggle-menu
manipulation = the act of purposely controlling, changing, or adapting one aspect or condition in an experiment
Anyone who has used a cellular phone knows that certain situations require a bit of research: If you suddenly find yourself in an area with poor phone reception, you might move a bit to the left or right, walk a few steps forward or back, or even hold the phone over your head to get a better signal. While the actions of a cell phone user might seem obvious, the person seeking cell phone reception is actually performing a scientific experiment: consciously manipulating one component (the location of the cell phone) and observing the effect of that action on another component (the phone's reception). Scientific experiments are obviously a bit more complicated, and generally involve more rigorous use of controls, but they draw on the same type of reasoning that we use in many everyday situations. In fact, the earliest documented scientific experiments were devised to answer a very common everyday question: how vision works.
A brief history of experimental methods
One of the first ideas regarding how human vision works came from the Greek philosopher Empedocles around 450 BCE. Empedocles reasoned that the Greek goddess Aphrodite had lit a fire in the human eye, and vision was possible because light rays from this fire emanated from the eye, illuminating objects around us. While a number of people challenged this proposal, the idea that light radiated from the human eye proved surprisingly persistent until around 1,000 CE, when a Middle Eastern scientist advanced our knowledge of the nature of light and, in so doing, developed a new and more rigorous approach to scientific research. Abū 'Alī al-Hasan ibn al-Hasan ibn al-Haytham, also known as Alhazen, was born in 965 CE in the Arabian city of Basra in what is now present day Iraq. He began his scientific studies in physics, mathematics, and other sciences after reading the works of several Greek philosophers. One of Alhazen's most significant contributions was a seven-volume work on optics titled Kitab al-Manazir (later translated to Latin as Opticae Thesaurus Alhazeni – Alhazen's Book of Optics). Beyond the contributions this book made to the field of optics, it was a remarkable work in that it based conclusions on experimental evidence rather than abstract reasoning – the first major publication to do so. Alhazen's contributions have proved so significant that his likeness was immortalized on the 2003 10,000-dinar note issued by Iraq (Figure 1).
Alhazen invested significant time studying light, color, shadows, rainbows, and other optical phenomena. Among this work was a study in which he stood in a darkened room with a small hole in one wall. Outside of the room, he hung two lanterns at different heights. Alhazen observed that the light from each lantern illuminated a different spot in the room, and each lighted spot formed a direct line with the hole and one of the lanterns outside the room. He also found that covering a lantern caused the spot it illuminated to darken, and exposing the lantern caused the spot to reappear. Thus, Alhazen provided some of the first experimental evidence that light does not emanate from the human eye but rather is emitted by certain objects (like lanterns) and travels from these objects in straight lines. Alhazen's experiment may seem simplistic today, but his methodology was groundbreaking: He developed a hypothesis based on observations of physical relationships (that light comes from objects), and then designed an experiment to test that hypothesis. Despite the simplicity of the method, Alhazen's experiment was a critical step in refuting the long-standing theory that light emanated from the human eye, and it was a major event in the development of modern scientific research methodology.
Experimentation as a scientific research method
Experimentation is one scientific research method, perhaps the most recognizable, in a spectrum of methods that also includes description, comparison, and modeling (see our Description, Comparison, and Modeling modules). While all of these methods share in common a scientific approach, experimentation is unique in that it involves the conscious manipulation of certain aspects of a real system and the observation of the effects of that manipulation. You could solve a cell phone reception problem by walking around a neighborhood until you see a cell phone tower, observing other cell phone users to see where those people who get the best reception are standing, or looking on the web for a map of cell phone signal coverage. All of these methods could also provide answers, but by moving around and testing reception yourself, you are experimenting.
Variables: Independent and dependent
In the experimental method, a condition or a parameter, generally referred to as a variable, is consciously manipulated (often referred to as a treatment) and the outcome or effect of that manipulation is observed on other variables. Variables are given different names depending on whether they are the ones manipulated or the ones observed:
Independent variable refers to a condition within an experiment that is manipulated by the scientist.
Dependent variable refers to an event or outcome of an experiment that might be affected by the manipulation of the independent variable.
Scientific experimentation helps to determine the nature of the relationship between independent and dependent variables. While it is often difficult, or sometimes impossible, to manipulate a single variable in an experiment, scientists often work to minimize the number of variables being manipulated. For example, as we move from one location to another to get better cell reception, we likely change the orientation of our body, perhaps from south-facing to east-facing, or we hold the cell phone at a different angle. Which variable affected reception: location, orientation, or angle of the phone? It is critical that scientists understand which aspects of their experiment they are manipulating so that they can accurately determine the impacts of that manipulation. In order to constrain the possible outcomes of an experimental procedure, most scientific experiments use a system of controls.
Controls: Negative, positive, and placebos
In a controlled study, a scientist essentially runs two (or more) parallel and simultaneous experiments: a treatment group, in which the effect of an experimental manipulation is observed on a dependent variable, and a control group, which uses all of the same conditions as the first with the exception of the actual treatment. Controls can fall into one of two groups: negative controls and positive controls.
In a negative control, the control group is exposed to all of the experimental conditions except for the actual treatment. The need to match all experimental conditions exactly is so great that, for example, in a trial for a new drug, the negative control group will be given a pill or liquid that looks exactly like the drug, except that it will not contain the drug itself, a control often referred to as a placebo. Negative controls allow scientists to measure the natural variability of the dependent variable(s), provide a means of measuring error in the experiment, and also provide a baseline to measure against the experimental treatment.
Some experimental designs also make use of positive controls. A positive control is run as a parallel experiment and generally involves the use of an alternative treatment that the researcher knows will have an effect on the dependent variable. For example, when testing the effectiveness of a new drug for pain relief, a scientist might administer treatment placebo to one group of patients as a negative control, and a known treatment like aspirin to a separate group of individuals as a positive control since the pain-relieving aspects of aspirin are well documented. In both cases, the controls allow scientists to quantify background variability and reject alternative hypotheses that might otherwise explain the effect of the treatment on the dependent variable.
Experimentation in practice: The case of Louis Pasteur
Well-controlled experiments generally provide strong evidence of causality, demonstrating whether the manipulation of one variable causes a response in another variable. For example, as early as the 6th century BCE, Anaximander, a Greek philosopher, speculated that life could be formed from a mixture of sea water, mud, and sunlight. The idea probably stemmed from the observation of worms, mosquitoes, and other insects "magically" appearing in mudflats and other shallow areas. While the suggestion was challenged on a number of occasions, the idea that living microorganisms could be spontaneously generated from air persisted until the middle of the 18th century.
In the 1750s, John Needham, a Scottish clergyman and naturalist, claimed to have proved that spontaneous generation does occur when he showed that microorganisms flourished in certain foods such as soup broth, even after they had been briefly boiled and covered. Several years later, the Italian abbot and biologist Lazzaro Spallanzani, boiled soup broth for over an hour and then placed bowls of this soup in different conditions, sealing some and leaving others exposed to air. Spallanzani found that microorganisms grew in the soup exposed to air but were absent from the sealed soup. He therefore challenged Needham's conclusions and hypothesized that microorganisms suspended in air settled onto the exposed soup but not the sealed soup, and rejected the idea of spontaneous generation.
Needham countered, arguing that the growth of bacteria in the soup was not due to microbes settling onto the soup from the air, but rather because spontaneous generation required contact with an intangible "life force" in the air itself. He proposed that Spallanzani's extensive boiling destroyed the "life force" present in the soup, preventing spontaneous generation in the sealed bowls but allowing air to replenish the life force in the open bowls. For several decades, scientists continued to debate the spontaneous generation theory of life, with support for the theory coming from several notable scientists including Félix Pouchet and Henry Bastion. Pouchet, Director of the Rouen Museum of Natural History in France, and Bastion, a well-known British bacteriologist, argued that living organisms could spontaneously arise from chemical processes such as fermentation and putrefaction. The debate became so heated that in 1860, the French Academy of Sciences established the Alhumbert prize of 2,500 francs to the first person who could conclusively resolve the conflict. In 1864, Louis Pasteur achieved that result with a series of well-controlled experiments and in doing so claimed the Alhumbert prize.
Pasteur prepared for his experiments by studying the work of others that came before him. In fact, in April 1861 Pasteur wrote to Pouchet to obtain a research description that Pouchet had published. In this letter, Pasteur writes:
Paris, 3 April 1861
The difference of our opinions on the famous question of spontaneous generation does not prevent me from esteeming highly your labor and praiseworthy efforts... The sincerity of these sentiments...permits me to have recourse to your obligingness in full confidence. I read with great care everything that you write on the subject that occupies both of us. Now, I cannot obtain a brochure that I understand you have just published.... I would be happy to have a copy of it because I am at present editing the totality of my observations, where naturally I criticize your assertions.
L. Pasteur (Porter, 1961)
Pasteur received the brochure from Pouchet several days later and went on to conduct his own experiments. In these, he repeated Spallanzani's method of boiling soup broth, but he divided the broth into portions and exposed these portions to different controlled conditions. Some broth was placed in flasks that had straight necks that were open to the air, some broth was placed in sealed flasks that were not open to the air, and some broth was placed into a specially designed set of swan-necked flasks, in which the broth would be open to the air but the air would have to travel a curved path before reaching the broth, thus preventing anything that might be present in the air from simply settling onto the soup (Figure 2). Pasteur then observed the response of the dependent variable (the growth of microorganisms) in response to the independent variable (the design of the flask). Pasteur's experiments contained both positive controls (samples in the straight necked flasks that he knew would become contaminated with microorganisms) and negative controls (samples in the sealed flasks that he knew would remain sterile). If spontaneous generation did indeed occur upon exposure to air, Pasteur hypothesized, microorganisms would be found in both the swan-neck flasks and the straight-necked flasks, but not in the sealed flasks. Instead, Pasteur found that microorganisms appeared in the straight necked flasks, but not in the sealed flasks or the swan-necked flasks.
By using controls and replicating his experiment (he used more than one of each type of flask), Pasteur was able to answer many of the questions that still surrounded the issue of spontaneous generation. Pasteur said of his experimental design, "I affirm with the most perfect sincerity that I have never had a single experiment, arranged as I have just explained, which gave me a doubtful result" (Porter, 1961). Pasteur's work helped refute the theory of spontaneous generation – his experiments showed that air alone was not the cause of bacterial growth in the flask, and his research supported the hypothesis that live microorganisms suspended in air could settle onto the broth in open-necked flasks via gravity.
Experimentation across disciplines
Experiments are used across all scientific disciplines to investigate a multitude of questions. In some cases, scientific experiments are used for exploratory purposes in which the scientist does not know what the dependent variable is. In this type of experiment, the scientist will manipulate an independent variable and observe what the effect of the manipulation is in order to identify a dependent variable (or variables). Exploratory experiments are sometimes used in nutritional biology when scientists probe the function and purpose of dietary nutrients. In one approach, a scientist will expose one group of animals to a normal diet, and a second group to a similar diet except that it is lacking a specific vitamin or nutrient. The researcher will then observe the two groups to see what specific physiological changes or medical problems arise in the group lacking the nutrient being studied.
Scientific experiments are also commonly used to quantify the magnitude of a relationship between two or more variables. For example, in the fields of pharmacology and toxicology, scientific experiments are used to determine the dose-response relationship of a new drug or chemical. In these approaches, researchers perform a series of experiments in which a population of organisms, such as laboratory mice, is separated into groups and each group is exposed to a different amount of the drug or chemical of interest. The analysis of the data that result from these experiments (see our Data Analysis and Interpretation module) involves comparing the degree of the organism's response to the dose of the substance administered.
In this context, experiments can provide additional evidence to complement other research methods. For example, in the 1950s a great debate ensued over whether or not the chemicals in cigarette smoke cause cancer. Several researchers had conducted comparative studies (see our Comparison in Scientific Research module) that indicated that patients who smoked had a higher probability of developing lung cancer when compared to nonsmokers. Comparative studies differ slightly from experimental methods in that you do not consciously manipulate a variable; rather you observe differences between two or more groups depending on whether or not they fall into a treatment or control group. Cigarette companies and lobbyists criticized these studies, suggesting that the relationship between smoking and lung cancer was coincidental. Several researchers noted the need for a clear dose-response study; however, the difficulties in getting cigarette smoke into the lungs of laboratory animals prevented this research. In the mid-1950s, Ernest Wynder and colleagues had an ingenious idea: They condensed the chemicals from cigarette smoke into a liquid and applied this in various doses to the skin of groups of mice. The researchers published data from a dose-response experiment of the effect of tobacco-smoke condensate on mice (Wynder et al., 1957).
As seen in Figure 3, the researchers found a positive relationship between the amount of condensate applied to the skin of mice and the number of cancers that developed. The graph shows the results of a study in which different groups of mice were exposed to increasing amounts of cigarette tar. The black dots indicate the percentage of each sample group of mice that developed cancer for a given amount cigarette smoke "condensate" applied to their skin. The vertical lines are error bars, showing the amount of uncertainty. The graph shows generally increasing cancer rates with greater exposure. This study was one of the first pieces of experimental evidence in the cigarette smoking debate, and it helped strengthen the case for cigarette smoke as the causative agent in lung cancer in smokers.
Sometimes experimental approaches and other research methods are not clearly distinct, or scientists may even use multiple research approaches in combination. For example, at 1:52 a.m. EDT on July 4, 2005, scientists with the National Aeronautics and Space Administration (NASA) conducted a study in which a 370 kg spacecraft named Deep Impact was purposely slammed into passing comet Tempel 1. A nearby spacecraft observed the impact and radioed data back to Earth. The research was partially descriptive in that it documented the chemical composition of the comet, but it was also partly experimental in that the effect of slamming the Deep Impact probe into the comet on the volatilization of previously undetected compounds, such as water, was assessed (A'Hearn et al., 2005). It is particularly common that experimentation and description overlap: Another example is Jane Goodall's research on the behavior of chimpanzees, which can be read in our Description in Scientific Research module.
Limitations of experimental methods
While scientific experiments provide invaluable data regarding causal relationships, they do have limitations. One criticism of experiments is that they do not necessarily represent real-world situations. In order to clearly identify the relationship between an independent variable and a dependent variable, experiments are designed so that many other contributing variables are fixed or eliminated. For example, in an experiment designed to quantify the effect of vitamin A dose on the metabolism of beta-carotene in humans, Shawna Lemke and colleagues had to precisely control the diet of their human volunteers (Lemke, Dueker et al. 2003). They asked their participants to limit their intake of foods rich in vitamin A and further asked that they maintain a precise log of all foods eaten for 1 week prior to their study. At the time of their study, they controlled their participants' diet by feeding them all the same meals, described in the methods section of their research article in this way:
Meals were controlled for time and content on the dose administration day. Lunch was served at 5.5 h postdosing and consisted of a frozen dinner (Enchiladas, Amy's Kitchen, Petaluma, CA), a blueberry bagel with jelly, 1 apple and 1 banana, and a large chocolate chunk cookie (Pepperidge Farm). Dinner was served 10.5 h post dose and consisted of a frozen dinner (Chinese Stir Fry, Amy's Kitchen) plus the bagel and fruit taken for lunch.
While this is an important aspect of making an experiment manageable and informative, it is often not representative of the real world, in which many variables may change at once, including the foods you eat. Still, experimental research is an excellent way of determining relationships between variables that can be later validated in real world settings through descriptive or comparative studies.
Design is critical to the success or failure of an experiment. Slight variations in the experimental set-up could strongly affect the outcome being measured. For example, during the 1950s, a number of experiments were conducted to evaluate the toxicity in mammals of the metal molybdenum, using rats as experimental subjects. Unexpectedly, these experiments seemed to indicate that the type of cage the rats were housed in affected the toxicity of molybdenum. In response, G. Brinkman and Russell Miller set up an experiment to investigate this observation (Brinkman & Miller, 1961). Brinkman and Miller fed two groups of rats a normal diet that was supplemented with 200 parts per million (ppm) of molybdenum. One group of rats was housed in galvanized steel (steel coated with zinc to reduce corrosion) cages and the second group was housed in stainless steel cages. Rats housed in the galvanized steel cages suffered more from molybdenum toxicity than the other group: They had higher concentrations of molybdenum in their livers and lower blood hemoglobin levels. It was then shown that when the rats chewed on their cages, those housed in the galvanized metal cages absorbed zinc plated onto the metal bars, and zinc is now known to affect the toxicity of molybdenum. In order to control for zinc exposure, then, stainless steel cages needed to be used for all rats.
Scientists also have an obligation to adhere to ethical limits in designing and conducting experiments. During World War II, doctors working in Nazi Germany conducted many heinous experiments using human subjects. Among them was an experiment meant to identify effective treatments for hypothermia in humans, in which concentration camp prisoners were forced to sit in ice water or left naked outdoors in freezing temperatures and then re-warmed by various means. Many of the exposed victims froze to death or suffered permanent injuries. As a result of the Nazi experiments and other unethical research, strict scientific ethical standards have been adopted by the United States and other governments, and by the scientific community at large. Among other things, ethical standards (see our Scientific Ethics module) require that the benefits of research outweigh the risks to human subjects, and those who participate do so voluntarily and only after they have been made fully aware of all the risks posed by the research. These guidelines have far-reaching effects: While the clearest indication of causation in the cigarette smoke and lung cancer debate would have been to design an experiment in which one group of people was asked to take up smoking and another group was asked to refrain from smoking, it would be highly unethical for a scientist to purposefully expose a group of healthy people to a suspected cancer causing agent. As an alternative, comparative studies (see our Comparison in Scientific Research module) were initiated in humans, and experimental studies focused on animal subjects. The combination of these and other studies provided even stronger evidence of the link between smoking and lung cancer than either one method alone would have.
Experimentation in modern practice
Like all scientific research, the results of experiments are shared with the scientific community, are built upon, and inspire additional experiments and research. For example, once Alhazen established that light given off by objects enters the human eye, the natural question that was asked was "What is the nature of light that enters the human eye?" Two common theories about the nature of light were debated for many years. Sir Isaac Newton was among the principal proponents of a theory suggesting that light was made of small particles. The English naturalist Robert Hooke (who held the interesting title of Curator of Experiments at the Royal Society of London) supported a different theory stating that light was a type of wave, like sound waves. In 1801, Thomas Young conducted a now classic scientific experiment that helped resolve this controversy. Young, like Alhazen, worked in a darkened room and allowed light to enter only through a small hole in a window shade (Figure 5). Young refocused the beam of light with mirrors and split the beam with a paper-thin card. The split light beams were then projected onto a screen, and formed an alternating light and dark banding pattern – that was a sign that light was indeed a wave (see our Light I: Particle or Wave? module).
Approximately 100 years later, in 1905, new experiments led Albert Einstein to conclude that light exhibits properties of both waves and particles. Einstein's dual wave-particle theory is now generally accepted by scientists.
Experiments continue to help refine our understanding of light even today. In addition to his wave-particle theory, Einstein also proposed that the speed of light was unchanging and absolute. Yet in 1998 a group of scientists led by Lene Hau showed that light could be slowed from its normal speed of 3 x 108 meters per second to a mere 17 meters per second with a special experimental apparatus (Hau et al., 1999). The series of experiments that began with Alhazen's work 1000 years ago has led to a progressively deeper understanding of the nature of light. Although the tools with which scientists conduct experiments may have become more complex, the principles behind controlled experiments are remarkably similar to those used by Pasteur and Alhazen hundreds of years ago.
Experimentation is a research method in which one or more variables are consciously manipulated and the outcome or effect of that manipulation on other variables is observed.
Experimental designs often make use of controls that provide a measure of variability within a system and a check for sources of error.
Experimental methods are commonly applied to determine causal relationships or to quantify the magnitude of response of a variable.
Agar, D. (2001). Arabic studies in physics and astronomy during 800-1400 AD. Retrieved August 1, 2007, from the University of Jyväskylä.
A'Hearn, M. F., Belton, M. J. S., Delamere, W. A., Kissel, J., Klaasen, K. P., McFadden, L. A., . . . White, R. L. (2005). Deep Impact: Excavating comet Tempel 1. Science, 310(5746), 258.
Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., & Walter, P. (2002). Molecular biology of the cell (4th ed.), London: Taylor & Francis.
Bajcsy, M., Zibrov, A. S., & Lukin, M. D. (2003). Stationary pulses of light in an atomic medium. Nature 426, 638-641.
Brinkman, G. L., & Miller, R. F. (1961). Influence of cage type and dietary zinc oxide upon molybdenum toxicity. Science, 134(3489), 1531.
Corbie-Smith, G. (1999). The continuing legacy of the Tuskegee Syphilis Study. American Journal of the Medical Sciences 317(1), 5-8.
Department of Health and Human Services. (Revised January 15, 2009). Code of Federal Regulations, 45 CFR §46, Protection of Human Subjects.
Fankhauser, D. B., & Stein Carter, J. (2004). Spontaneous generation. Retrieved August 1, 2007, from the Clermont College Biology Department.
Gorini, R. (2003). Al-Haytham the man of experience. First steps in the science of vision. Journal of the International Society for the History of Islamic Medicine, 2(4), 53-55.
Hau, L. V., Harris, S. E., Dutton, Z., & Behroozi, C. H. (1999). Light speed reduction to 17 metres per second in an ultracold atomic gas. Nature, 397, 594-598.
Kent, J. (2006, January).The Impact of the scientific revolution: A brief history of the experimental method in the 17th century. Retrieved August 1, 2007, from Connexions.
Lemke S. L., Dueker, S. R., Follett, J. R., Lin, Y., Carkeet, C., Buchholz, B. A., Vogel, J. S., & Clifford, A. J.(2003). Absorption and retinol equivalence of beta-carotene in humans is influenced by dietary vitamin A intake. Journal of Lipid Research, 44(8), 1591-600.
O'Connor, J. J., & Robertson, E. F. (2002). Light through the ages: Ancient Greece to Maxwell. Retrieved August 1, 2007, from the MacTutor History of Mathematics Archive, School of Mathematics and Statistics, University of St. Andrews, Scotland.
O'Connor, J. J., & Robertson, E. F. (1999). Abu Ali al-Hasan ibn al-Haytham. Retrieved August 1, 2007, from the MacTutor History of Mathematics Archive, School of Mathematics and Statistics, University of St. Andrews, Scotland.
Pasteur, L. (1861). Sur les corpuscules organizes qui existent dans l'atmosphère. Examen de la doctrine des générations spontanées. (Legon professée á la Societé chimique de Paris, le 19 mai). Mémoire sur les corpuscules organisés qui existent en suspension dans l'atmosphère. Examen de la doctrine des générations spontanées (Extrait). Comptes rendus, 52(1), 1142-1143, 3 juin. Mémoire sur les corpuscules organisés qui existent dans l'atmosphère. Examen de la doctrine des générations spontanées. Annales des sciences naturelles (partie zoologique) (Sér. 4) 16, 5-98.
Porter, J. R. (1961). Louis Pasteur: Achievements and disappointments, 1861. Bacteriology Reviews, 25(4), 389–403.
Reuben, A. (2004). Au Conntraire, professeur Pasteur! Hepatology 40(6), 1478-1482.
Wynder, E. L., Kopf, P., & Ziegler, H. (1957). A study of tobacco carcinogenesis: II. Dose-response studies. Cancer 10(6).