Not sure which lesson to do first? Find your level and get a plan →

Math · Statistics & Probability · Grade 6-8

Observational vs Experimental Studies

An observational study records data without imposing a treatment, while an experiment deliberately manipulates a variable; only an experiment with random assignment can establish causation.


Quick Answer

An observational study records data without imposing a treatment, while an experiment deliberately manipulates a variable; only an experiment with random assignment can establish causation. Use the distinction whenever you read a claim like "coffee drinkers live longer" or "students who eat breakfast score higher": first find out whether the researchers assigned the treatment or only recorded what people already did. That decides whether the study supports "causes" or only "is linked to," and so how far its conclusion can go.

Why This Matters

News stories turn associations into advice every day: eat this, live longer. Many of those claims come from observational studies, where the people who chose the habit also differ in income, exercise or health. Knowing the study type tells you whether a confounder could explain the result, and whether changing your own habit would actually change your outcome.

Intuitive Explanation

A study finds that students who eat breakfast score higher on tests. Does breakfast cause better scores? The students chose for themselves whether to eat breakfast, so this is an observational study. Breakfast eaters might also sleep more or have parents who help with homework. Any of these lurking variables, called confounders, could produce the higher scores, with breakfast simply along for the ride.

Now imagine a school randomly assigns 100 volunteers: 50 get a school breakfast and 50 don't. Random assignment spreads sleep habits, family support and everything else roughly evenly across the two groups. If the breakfast group then scores higher, the only systematic difference left is breakfast. That is why an experiment can say "causes" while an observational study can only say "is associated with."

Sometimes an experiment is impossible or wrong. Nobody can randomly assign people to smoke for 30 years, so the link between smoking and lung cancer came from observational studies: many of them, in different countries, showing that heavier smokers got sick more often, backed by lab evidence of how smoke damages cells. Strong observational evidence can be convincing, but each single study still shows only association.

Two groups do not make an experiment. Comparing students who chose band with students who didn't is observational, even with equal group sizes and a huge sample; a bigger sample shrinks random error but does not remove confounders.

Think of it like this. Counting umbrellas on a street and noticing that wet pavement comes with them might suggest umbrellas wet the ground. Rain causes both, a hidden common cause, just like a confounder. An experiment would hand out umbrellas at random on a dry day and check the pavement, which breaks the link to rain and shows umbrellas do nothing to the ground.

Core idea. Observational studies record what already happens (association only); experiments assign treatments and can establish causation.


When to Use

Use this distinction when a question asks whether a study proves that one thing causes another, or whether a conclusion is justified. Find out how the groups formed: if subjects were randomly assigned, causation can be claimed; if they chose the treatment or already had the trait, only association can. Then look for confounders. If the question is about how participants were selected from a population, that is a sampling method, which affects whom the results apply to rather than whether they show cause.

Tip. Ask: Did the researcher assign the treatment to subjects, or only observe a group that already had it?

How to Recognize It

Before you judge what a study proves, look at how its groups came to be and how strong a claim its conclusion makes.

  1. Is the factor being studied something nobody could hand out, such as a person's age, height or the town they grew up in?

    If yes, the study must be observational, since no one can assign those traits, and only a link can be reported. If the factor can be given, such as a drink or a teaching method, check whether it was handed out at random.

  2. Does the conclusion claim that one thing produces a change in another, rather than that the two go together?

    If yes, check that the study was a randomized experiment; an observational study only supports phrases like linked to or associated with. If the claim is already only about a link, the study type does not limit it.

  3. Can you think of a third factor that could push both variables up or down together?

    If yes, that factor is a possible confounder, as rain is for umbrellas and wet pavement. Random assignment would spread it evenly across the groups; without it, the whole link may come from the confounder.

  4. Is the complaint really about who was chosen to take part, such as only people who answered an online poll?

    If yes, that is sampling bias, a problem with selection, and it can damage an experiment or an observational study alike. It limits whom the results apply to, not whether they show cause.

Observational vs Experimental Studies compared with Experimental design, Correlation vs causation, and Sampling bias

Each row judges what a study may claim, so they get mixed up. Ask what is judged: the study type, which caps the claim (this row); the plan inside an experiment (design); what a relationship means (correlation vs causation); or how subjects were picked (sampling bias).

Observational vs Experimental Studies

Meaning
Use this when you must decide whether a study can claim causation or only association, by checking if treatments were assigned.
Key test
Did the researcher assign the treatment to subjects, or only observe a group that already had it?
Example
A study follows 5,000 adults for 20 years, finds coffee drinkers live longer, and concludes coffee extends life. Is the conclusion valid?

Experimental design

Meaning
The detailed plan (control, randomization, blinding) inside an experiment, not the watch-vs-assign distinction.
Key test
Use when building the experiment's structure, not classifying study type.
Example
Choosing how to randomize and blind

Correlation vs causation

Meaning
A statement about variables moving together vs one causing the other; this concept is about study TYPE that licenses each claim.
Key test
Use when interpreting a relationship rather than classifying the study.
Example
Ice-cream sales and drownings both rise in summer

Sampling bias

Meaning
A flaw in HOW subjects were chosen, which can wreck either study type.
Key test
Use when the concern is an unrepresentative sample, not the cause-vs-association ceiling.
Example
Surveying only gym members about exercise

Formula & Notation

Worked Examples

Example 1 — Coffee and longevity

Easy

A study follows 5,000 adults for 20 years, finds coffee drinkers live longer, and concludes coffee extends life. Is the conclusion valid?
Show the steps and the answer
  1. Ask how people ended up in the coffee group: they chose to drink coffee.

    The researchers assigned nothing, so this is an observational study.

  2. Name possible confounders: coffee drinkers may be wealthier or have better health care, and some non-drinkers may have given up coffee because they were already ill.

    Each of these could lengthen or shorten life on its own, apart from coffee.

  3. Note that 5,000 people and 20 years don't fix this.

    A large, long study measures the association precisely, but the confounders are still mixed into it.

  4. Restate the valid conclusion: coffee drinking is associated with longer life in this group.

    Claiming that coffee extends life would need random assignment to rule out the confounders.

Answer: Only association is justified, not causation

Takeaway: Self-chosen groups allow only 'linked to', however large the study.

Example 2 — Randomized coffee trial

Standard

Now researchers randomly assign 5,000 volunteers to drink coffee or not, then track survival. Can this claim causation?
Show the steps and the answer
  1. Ask how people ended up in each group: a random draw decided.

    The researchers imposed the treatment, so this is an experiment.

  2. Revisit the earlier confounders: wealth, health care, existing illness.

    With about 2,500 people per group chosen by chance, these end up roughly balanced between the groups.

  3. Conclude that a clear survival difference can be credited to coffee, for people like these volunteers.

    Volunteers may differ from the general public, so the causal claim applies most safely to people like them.

Answer: Yes — random assignment supports a causal claim

Takeaway: Random assignment balances confounders, which is what licenses 'causes'; who volunteered limits whom the result applies to.

Example 3 — Find the mistake: two groups are not an experiment

Application

Researchers compare 400 students who chose to take music lessons with 400 who did not. The music group's average math score is 78 and the other group's is 72. Zoe writes: "There are two equal groups and a big sample, so this is an experiment, and music lessons raise math scores by 6 points." Find the mistake and correct it.
Show the steps and the answer
  1. Ask who decided which group each student was in: the students and their families did.

    Nothing was assigned, so the study is observational whatever the group sizes.

  2. Find confounders: families who pay for music lessons may also pay for tutoring, and students who stick with lessons may have stronger study habits.

    Either could raise math scores with no help from music.

  3. Check what the big sample does achieve: with 400 per group, a gap of 78−72=6 points is unlikely to be pure chance.

    Size makes the association reliable, but it can't separate music from the confounders that come with it.

  4. State the supported claim: music students scored 6 points higher on average.

    That is an association; showing cause would require randomly assigning students to lessons.

Answer: This is an observational study: music lessons are associated with 6-point higher math scores, but the study can't show that the lessons caused the gap.

Takeaway: Equal group sizes and big samples don't make an experiment; only random assignment of the treatment does.

Where You'll See It

  • Smoking and lung cancer

    No one can ethically assign people to smoke, so the evidence came from observational studies. Scientists became convinced the link was causal because many studies agreed, heavier smokers had higher cancer rates, and lab work showed how smoke damages lung cells, not because any single study proved it.

  • The hormone therapy reversal

    Observational studies once suggested that hormone therapy protected women from heart disease. A large randomized trial reported in 2002 found no such protection and some added risk. The women who had chosen hormone therapy tended to be healthier and wealthier to begin with.

  • Screen time and sleep

    A survey finds that teens with more screen time report less sleep. Since teens chose their own screen time, the link could run either way (poor sleepers may reach for their phones at night) or come from a third factor such as stress. Only an experiment that randomly limited screen time could show cause.

Common Mistakes

  1. The slip: Claiming an observational study proved causation

    Instead: without random assignment, only association can be claimed.

  2. The slip: Thinking 'large sample' upgrades an observational study to causal

    Instead: size doesn't remove confounders; only random assignment does.

  3. The slip: Calling any study with two groups an experiment

    Instead: if subjects already belonged to their group, it's observational.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. Researchers randomly split 80 tomato plants into two groups of 40, give one group extra light each day, and compare the harvests. Observational study or experiment? Can it show cause?

    Hint: Did the researchers impose the treatment, and how did they form the groups?

  2. A survey of 2,000 teens finds that those who sleep more than 8 hours get higher grades. What conclusion is justified, and what lurking variable could explain it?

    Hint: Did anyone assign the amount of sleep?

  3. An observational study found that kids who read on paper remember more than kids who read on screens. Describe an experiment with 60 students that could test cause.

    Hint: Let chance decide who reads on paper.

  4. A study compares 40 students who already chose online classes with 40 who chose in-person classes. There are two groups. Is this an experiment?

    Hint: Who placed each student in a group?

  5. A study of 100,000 adults finds dog owners exercise more. Ana says: "With 100,000 people, this proves owning a dog makes you exercise more." Correct Ana.

    Hint: Does a large sample remove the reasons people chose to own dogs?

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

If observational studies cannot prove cause, how do we know smoking causes cancer?

No one would ever assign people to smoke, so the human evidence is observational, but it is very strong. Many studies in different countries found the same link; heavier smokers got cancer more often; quitting lowered the risk; and lab experiments on animals and cells showed how smoke damages tissue. When many kinds of evidence point the same way and no confounder explains them all, scientists accept the cause.

Is a survey an observational study?

Yes. A survey asks people questions and records their answers without giving them any treatment, so it is a kind of observational study. A survey can show that students who read more also report better sleep, but it cannot show that reading causes better sleep. Surveys are excellent for describing what a group thinks or does, as long as the people asked were chosen fairly.

Can an observational study ever be the better choice?

Sometimes it is the only honest option. You cannot randomly assign people to be born in a certain year, to live near a highway or to lose a job. Observational studies can also follow huge groups for decades, which experiments rarely do, and they watch people in real life rather than in a lab. The cost is that their conclusions stay at the level of association.

What is reverse causation?

It means cause and effect might run the other way. Suppose a study finds that people who exercise more are happier. Exercise might make people happy, but happier people might also have more energy to exercise. An observational study cannot tell which direction is true. An experiment that randomly assigns some people to an exercise program settles it, because the researcher, not the person's mood, decided who exercised.

Learning Path

Observational vs Experimental Studies

You are here

Next →

Nothing after this yet
You need Experimental Design, Causation and Correlation first. This concept adds one sorting decision for any study (assigned or observed) and the claim each type supports: cause or association. With it you can judge news headlines and research claims, spot confounders, and say what further evidence, such as a randomized trial, would settle a question; later, in statistical inference, the study type limits every conclusion you draw.

See Also