# How do experiments test media effects?

> Written with AI · [Communications: research methods in communication](https://lessoons.com/en/communication-research-methods) · Grade 12 · 2026-11-14 · https://lessoons.com/en/communication-research-methods/how-do-experiments-test-media-effects

An experiment tests media effects by randomly dividing people into groups, exposing only one group to a message, and then measuring both groups in the same way. Because random assignment makes the groups alike in everything else, a difference at the end can be credited to the message. That is something a survey of viewing habits cannot show, since it finds links but not causes.

## The topic explained

### The question only an experiment can answer

Media researchers often want to know whether a message causes a change: does a news story make a problem seem more important, does an ad change attitudes, does a violent game make players more aggressive? A survey can show that people who watch more news on a subject care more about it, but that is only a link. Perhaps people who already care about the subject choose to watch more. An experiment is built to cut through that problem.

The researcher changes one thing, the independent variable, such as which version of a newscast people see, and measures the outcome, the dependent variable, such as how important they rate a problem. The key step is random assignment: a coin toss or a computer decides who goes into which group. With enough participants, randomness spreads age, interest, politics and mood evenly, so the groups differ only in the message. The group that gets the message is the treatment group; the group that does not is the control group.

### A famous experiment with the evening news

In 1982 Shanto Iyengar, Mark Peters and Donald Kinder published experiments in the American Political Science Review. Volunteers watched evening newscasts that the researchers had quietly edited, so that some groups saw extra stories about one national problem, such as weaknesses in defense. Later, the participants who had seen those stories rated that problem as more important than participants who had not, and they gave it more weight when judging the president. Agenda-setting researchers, following McCombs and Shaw's 1972 study, had found this link in surveys; the experiment showed that the news itself produced it, and it added a new idea, priming: the news changes the yardstick people use to judge leaders. Iyengar and Kinder expanded the work in the book News That Matters in 1987.

Here is a made-up version with numbers. 120 students are split at random into two groups of 60. The treatment group watches a newscast with four extra reports on road accidents; the control group watches the same newscast without them. Then everyone is asked which problem is the most important. 27 of 60 in the treatment group name road safety, which is 45%, against 9 of 60 in the control group, 15%. The difference, 30 percentage points, is the estimated effect of the extra reports.

### Strengths, limits and ethics

An experiment is strong on internal validity: its conclusion about cause is sound. Its weakness is external validity, whether the result holds outside the study. A short lab session with a single message is not the same as months of real viewing at home, and volunteers are often students rather than a cross-section of the public. That is why researchers combine experiments with surveys and field studies. A field experiment moves the design into real life: a news site, for instance, can show two versions of a page to readers chosen at random and compare what they read next. It is more realistic, but the researcher controls less of what else happens to the participants.

Experiments that hide their real purpose must protect participants: taking part is voluntary, answers stay confidential, and at the end the researcher holds a debriefing, explaining what was changed and why. A common exam mistake is to call any study with two groups an experiment. If people chose their own group, for example heavy and light news viewers, it is a comparison, not an experiment, and it cannot show cause.

## Practice: experiments in media research

For 12th grade communications students. Every study, group and number here is made up unless the page names a real study. Answer from what is printed on the page, then check yourself against the answers.

1. **What is the missing step in the design of the experiment, and why is it essential?** — Random split (random assignment). It makes the groups alike in everything except the newscast, so a difference at the end can be credited to the newscast.
2. **In the made-up experiment, what is the independent variable and what is the dependent variable?** — Independent: which newscast the group watched (with or without the extra road reports). Dependent: the problem participants named as the most important.
3. **Use the chart. How many percentage points separate the groups, and what does that number estimate?** — 45% − 15% = 30 percentage points. It estimates the effect of the extra reports on how important road safety seems.
4. **Fill in the missing cell of the table.** — No, only a link: a survey finds that two things go together, but not which causes which.
5. **A made-up study compares 200 teenagers who chose to play violent games with 200 who chose not to, and finds the first group more aggressive. Is this an experiment? Can it show that the games cause aggression?** — No. The teenagers chose their own group, so the groups may differ in other ways, for example aggressive teens may prefer violent games. It shows a link, not a cause.
6. **What did Iyengar, Peters and Kinder's experiments of 1982 show?** — That people who watched newscasts edited to include more stories about a problem rated that problem as more important and gave it more weight in judging the president: the news caused the change.
7. **Give one reason the result of a lab experiment might not hold in real life, and one duty the researcher has toward participants.** — Reason: a short session with one message is not like months of viewing at home, or the volunteers are not typical of the public. Duty: voluntary participation, confidentiality, or a debriefing that explains what was changed.

## Common questions

### Why is random assignment so important?

Because it makes the groups alike in age, interests, opinions and everything else, so that the only systematic difference between them is the message. Without it, a difference at the end could come from who chose which group.

### What is a control group?

The group in an experiment that does not get the message, or gets a neutral version. Comparing it with the treatment group shows how much of the outcome is due to the message.

### What is the difference between an experiment and a survey in media research?

An experiment changes one thing and measures the effect, so it can show cause. A survey asks people about what they already do and think, so it can show links between things but not which causes which.

### What are internal and external validity?

Internal validity is whether a study's conclusion about cause is sound. External validity is whether the result holds outside the study, with other people and in real life.

### Who used experiments to study television news?

Shanto Iyengar, Mark Peters and Donald Kinder, in experiments published in 1982, and Iyengar and Kinder in the 1987 book News That Matters, showed that edited newscasts changed which problems viewers considered important.
