If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

Main content

Controlled experiments

How scientists conduct experiments and make observations to test hypotheses.


Biologists and other scientists use the scientific method to ask questions about the natural world. The scientific method begins with an observation, which leads the scientist to ask a question. They then come up with a hypothesis, a testable explanation that addresses the question.
A hypothesis isn't necessarily right. Instead, it's a "best guess," and the scientist must test it to see if it's actually correct. Scientists test hypotheses by making predictions: if hypothesis X is right, then Y should be true. Then, they do experiments or make observations to see if the predictions are correct. If they are, the hypothesis is supported. If they aren't, it may be time for a new hypothesis.

How are hypotheses tested?

When possible, scientists test their hypotheses using controlled experiments. A controlled experiment is a scientific test done under controlled conditions, meaning that just one (or a few) factors are changed at a time, while all others are kept constant. We'll look closely at controlled experiments in the next section.
In some cases, there is no good way to test a hypothesis using a controlled experiment (for practical or ethical reasons). In that case, a scientist may test a hypothesis by making predictions about patterns that should be seen in nature if the hypothesis is correct. Then, they can collect data to see if the pattern is actually there.

Controlled experiments

What are the key ingredients of a controlled experiment? To illustrate, let's consider a simple (even silly) example.
Suppose I decide to grow bean sprouts in my kitchen, near the window. I put bean seeds in a pot with soil, set them on the windowsill, and wait for them to sprout. However, after several weeks, I have no sprouts. Why not? Well...it turns out I forgot to water the seeds. So, I hypothesize that they didn't sprout due to lack of water.
To test my hypothesis, I do a controlled experiment. In this experiment, I set up two identical pots. Both contain ten bean seeds planted in the same type of soil, and both are placed in the same window. In fact, there is only one thing that I do differently to the two pots:
  • One pot of seeds gets watered every afternoon.
  • The other pot of seeds doesn't get any water at all.
After a week, nine out of ten seeds in the watered pot have sprouted, while none of the seeds in the dry pot have sprouted. It looks like the "seeds need water" hypothesis is probably correct!
Let's see how this simple example illustrates the parts of a controlled experiment.
Panel 1: Two identical pots are prepared. 10 bean seeds are added to each pot. The pots are placed near the window.
Panel 2: One pot (experimental group) is watered. The other pot (control group) is not watered. The independent variable is the amount of water given.
Panel 3: In the experimental (watered) pot, 9/10 seed sprout. In the control (unwatered) pot, 0/10 seeds sprout. The fraction of seeds that sprout is the dependent variable.

Control and experimental groups

There are two groups in the experiment, and they are identical except that one receives a treatment (water) while the other does not. The group that receives the treatment in an experiment (here, the watered pot) is called the experimental group, while the group that does not receive the treatment (here, the dry pot) is called the control group. The control group provides a baseline that lets us see if the treatment has an effect.

Independent and dependent variables

The factor that is different between the control and experimental groups (in this case, the amount of water) is known as the independent variable. This variable is independent because it does not depend on what happens in the experiment. Instead, it is something that the experimenter applies or chooses him/herself.
In contrast, the dependent variable in an experiment is the response that's measured to see if the treatment had an effect. In this case, the fraction of bean seeds that sprouted is the dependent variable. The dependent variable (fraction of seeds sprouting) depends on the independent variable (the amount of water), and not vice versa.
Experimental data (singular: datum) are observations made during the experiment. In this case, the data we collected were the number of bean sprouts in each pot after a week.

Variability and repetition

Out of the ten watered bean seeds, only nine came up. What happened to the tenth seed? That seed may have been dead, unhealthy, or just slow to sprout. Especially in biology (which studies complex, living things), there is often variation in the material used for an experiment – here, the bean seeds – that the experimenter cannot see.
Because of this potential for variation, biology experiments need to have a large sample size and, ideally, be repeated several times. Sample size refers to the number of individual items tested in an experiment – in this case, 10 bean seeds per group. Having more samples and repeating the experiment more times makes it less likely that we will reach a wrong conclusion because of random variation.
Biologists and other scientists also use statistical tests to help them distinguish real differences from differences due to random variation (e.g., when comparing experimental and control groups).

Controlled experiment case study: CO2 and coral bleaching

As a more realistic example of a controlled experiment, let's examine a recent study on coral bleaching. Corals normally have tiny photosynthetic organisms living inside of them, and bleaching happens when they leave the coral, typically due to environmental stress. The photo below shows a bleached coral in front and a healthy coral in back.
Photograph showing a bleached, white coral in the foreground and a healthy, brownish coral in the background.
Image credit: "Keppelbleaching" (CC BY 3.0).
A lot of research on the cause of bleaching has focused on water temperature1. However, a team of Australian researchers hypothesized that other factors might be important too. Specifically, they tested the hypothesis that high CO2 levels, which make ocean waters more acidic, might also promote bleaching2.
What kind of experiment would you do to test this hypothesis? Think about:
  • What your control and experimental groups would be
  • What your independent and dependent variables would be
  • What results you would predict in each group

Non-experimental hypothesis tests

Some types of hypotheses can't be tested in controlled experiments for ethical or practical reasons. For example, a hypothesis about viral infection can't be tested by dividing healthy people into two groups and infecting one group: infecting healthy people would not be safe or ethical. Similarly, an ecologist studying the effects of rainfall can't make it rain in one part of a continent, while keeping another part dry as a control.
In situations like these, biologists may use non-experimental forms of hypothesis testing. In a non-experimental hypothesis test, a researcher predicts observations or patterns that should be seen in nature if the hypothesis is correct. They then collect and analyze data, seeing whether the patterns are actually present.

Case study: Coral bleaching and temperature

A good example of hypothesis testing based on observation comes from early studies of coral bleaching. As mentioned above, bleaching is when corals lose the photosynthetic microorganisms that live inside of them, which makes them turn white. Researchers suspected that high water temperature might cause bleaching, and they tested this hypothesis experimentally on a small scale (using isolated coral fragments in tanks)3,4.
What ecologists most wanted to know, however, was whether water temperature was causing bleaching for lots of different coral species in their natural setting. This broader question could not be answered experimentally, as it wouldn't be ethical (or even possible) to artificially change the water temperature surrounding entire coral reefs.
False-colored map representing sea surface temperatures around the globe as different colors. Warmers colors, mostly near the equator, represent hotter temperatures, while cooler colors, mostly near the poles, represent cooler temperatures.
Image credit: "Global sea surface temperature," by NASA (public domain).
Instead, to test the hypothesis that natural bleaching events were caused by increases in water temperature, a team of researchers wrote a computer program to predict bleaching events based on real-time water temperature data. For example, this program would generally predict bleaching for a particular reef when the water temperature in the reef's area exceeded its average monthly maximum by 1 C or more1.
The computer program was able to predict many bleaching events weeks or even months before they were reported, including a large bleaching event in the Great Barrier Reef in 19981. The fact that a temperature-based model could predict bleaching events supported the hypothesis that high water temperature causes bleaching in naturally occurring coral reefs.

Want to join the conversation?

  • aqualine ultimate style avatar for user Yvonne
    In the case of environmental effects on children and how it affects their health, it being unethical to actually try an experiment, then from what I gather the best way to test a hypothesis would be to say, "When children are exposed to X, Y will happen." and then to collect data. Would legitimate studies only research this from children that are enrolled in such a study at birth with doctors/scientists/researchers taking blood samples or personally investigating the home environment for these factors? Or could this data be reported by parents? How would you account for people who are dishonest on their reports (as Gregory House states in nearly every episode, "Everybody lies.")? Is a retroactive study considered legitimate if children aren't entered into the study until they are older in age? Or if the researcher only relies on the parents' reporting data from birth to the current age that the child is involved in testing a hypothesis?
    (83 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Companion1
    when the coral reefs are bleached does that mean they have died, if not do they stay white forever or return to their original form?
    (26 votes)
    Default Khan Academy avatar avatar for user
    • duskpin ultimate style avatar for user Herta Elīna Jansone
      When a coral is bleaching, it expels symbiotic algae (zooxanthellae), which supply food for the coral, get rid of metabolic byproducts, protect it from pathogens and also create its bright colors. Bleaching results in starvation, which is indicated by loss of color. This doesn't mean that they have died - if the corals find a different nutrient-supplying species, they can regenerate and stabilize again. If they don't find new symbionts and the factors which contribute to bleaching persist, then the coral might die of starvation or disease.
      A colony of corals might regenerate if corals or their polyps from other colonies migrate and reproduce successfully elsewhere, where the conditions for growth are more favorable.
      (46 votes)
  • duskpin ultimate style avatar for user Jack Smith
    So hypotheses are like conditional statements?

    Thanks to anyone who answers,
    (21 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user kelly smith
    why not test different acidity levels with different temperatures?
    (11 votes)
    Default Khan Academy avatar avatar for user
    • piceratops ultimate style avatar for user Colin Andrews
      If each one had different temperatures, then you have too many variables. Let's say you did the same experiment as the Australian researchers, but also had each sample at a different temperature. Let's say you get the following results:

      Low acidity with cold temperature: 10% bleached

      Medium acidity with mild temperature: 30% bleached

      High acidity with hot temperature: 50% bleached

      Now that you have your results, you know what causes bleaching in coral. But, do you? Is it temperature that causes it? Is it the acidity of the water? Is it the combination of both? How can you tell? This causes confusion and possibly false claims. If you want to know both, it is better to have two seperate tests. One for "Is Acidity Causing Coral Bleach" and another for "Is Temperature Causing Coral Bleach".

      At the end, you will have two results, two claims, and zero variable confusion.

      I hope this answers your question!

      (18 votes)
  • female robot grace style avatar for user debwallace74
    How many times do you need to repeat an experiment in order to see if your hypothesis is nearly true or not?
    Also how big does the sample size need to be to conduct the experiment? 10,20,100?
    (11 votes)
    Default Khan Academy avatar avatar for user
    • winston baby style avatar for user Ivana - Science trainee
      There are no official rules about how many times experiments should be repeated to be reliable. However, the replication level of your experiments will have an impact on the statistical tests you can perform on your dataset (parametric or non-parametric tests, which do not have the same power).

      If the aim is to determine whether an event has occurred (for example, whether a pathogen is present in a colony of animals), the number of animals that need to be tested or produced is given by

      Image p20007f55g177001.jpg
      where β is the probability of committing a Type II error (usually 0.10 or 0.05) and p represents the proportion of the animals in the colony that are not infected.

      *Sample size depends on type of experiment.*

      Here is paper with formula:
      (13 votes)
  • sneak peak green style avatar for user 小苹果
    What is another example of an everyday use of controlled experiments?
    (11 votes)
    Default Khan Academy avatar avatar for user
  • mr pants orange style avatar for user Brian F
    How do the sprouting seeds know to go up if they don't know which way the light is?
    (10 votes)
    Default Khan Academy avatar avatar for user
    • blobby green style avatar for user madhoolika.chodavarapu
      Plants are PHOTOTROPIC I.e, they sense light and act accordingly. Some plants are negatively phototropic ( go against light) while others are positively phototropic (go towards light).

      There is also GEOTROPISM, there plants that are negatively geotropic, go against gravity (upwards) while some others are positively geotropic ( the go towards the gravity).

      These phenomenons come under the topic TROPISM
      (5 votes)
  • blobby green style avatar for user lefilsdarren
    Isn't taking coral reefs from their environment unethical since corals are disappearing? Why not create an experiment that would remove co2 levels by adding plants that sequester carbon near the bleached corols? The corols in the water that do not have plants that sequester carbon would be the control group, while the others would be the experimental groups. The independent variables would either be the co2 or pH levels. The dependent variable would be the degree of bleaching from the corals. You would do a side-by-side comparison of each of the waters that have different co2 levels. Then, you could do large/repeated sample sizes by testing waters with high/medium/regular co2 levels that have corals in the water.
    (11 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user 47charbrow
    how come a prediction is not the same thing as a hypothesis? isn't a hypothesis what your predicting will happen?
    (7 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Asithandile Ntsondwa
    in the results obtained how does the subtraction of the control affect the results and why
    (7 votes)
    Default Khan Academy avatar avatar for user