Study smarter with Fiveable
Get study guides, practice questions, and cheatsheets for all your subjects. Join 500,000+ students with a 96% pass rate.
Experimental design is the backbone of causal research in communication studies. It's how we move beyond "these two things are related" to "this actually causes that." You're being tested on your ability to distinguish between design types, understand their trade-offs, and recommend the right approach for different research scenarios. The core tensions you'll encounter repeatedly are internal validity vs. external validity, control vs. realism, and statistical power vs. practical constraints.
Don't just memorize design names. Know what problem each design solves, what it sacrifices in the process, and when a researcher would choose one over another. FRQs love asking you to justify design choices or identify threats to validity, so think like a researcher making strategic decisions, not a student reciting definitions.
The most fundamental distinction in experimental research is how much control the researcher has over participant assignment. This single factor determines whether you can make causal claims with confidence or must hedge your conclusions.
Random assignment is the defining feature. Every participant has an equal chance of being placed in any condition, which distributes confounding variables (like personality, prior knowledge, or motivation) roughly evenly across groups. Because of this, the independent variable becomes the only systematic difference between groups, allowing the researcher to establish cause-and-effect relationships.
True experiments maximize internal validity, but they may sacrifice external validity if the controlled conditions feel artificial to participants. A tightly controlled lab study on persuasion, for instance, might not reflect how people actually process messages in their daily lives.
Quasi-experiments use pre-existing groups instead of random assignment. For example, you might compare communication styles across different university departments or study a training program's effects on employees who already belong to separate teams.
This design is essential for real-world research where randomization is unethical, impractical, or impossible. The trade-off is that selection bias becomes the primary threat: differences between groups may exist before the study even begins, which weakens any causal claims you try to make.
Compare: True experimental vs. quasi-experimental design: both manipulate an independent variable, but only true experiments use random assignment. If an FRQ asks about studying workplace communication interventions across existing teams, quasi-experimental is your answer. If it asks about maximizing causal certainty, go with true experimental.
How and when you measure participants shapes what conclusions you can draw. Pre-testing can strengthen your analysis but may also contaminate your results.
This design measures participants both before and after the intervention. The pre-test establishes a baseline so you can see how much each group actually changed. A control group receives no treatment, which lets you attribute post-test differences to the intervention rather than to external factors like current events or natural maturation.
The risk? The pre-test itself might sensitize participants. If you give people a questionnaire about their media habits and then expose them to a media literacy intervention, they might pay closer attention to the treatment simply because the pre-test made them think about the topic.
This design directly addresses the sensitization problem by adding two groups that skip the pre-test entirely. That gives you four groups total:
By comparing outcomes across all four groups, you can determine whether participants changed because of the treatment itself or because the pre-test primed them to respond differently. It's the gold standard for controlling pre-test effects, but it requires roughly double the participants and resources of a standard pre-test/post-test design.
Compare: Pre-test/post-test vs. Solomon four-group: both use control groups, but Solomon specifically isolates pre-test sensitization effects. When an exam question mentions concerns about testing effects or reactivity, Solomon is the design that addresses this.
The decision about whether participants experience one condition or multiple conditions affects everything from sample size requirements to the types of bias you'll encounter.
Each participant experiences only one condition. Group A sees persuasive message version 1, Group B sees version 2, and the two groups never cross over. This completely eliminates carryover effects since participants can't be influenced by prior exposure to other conditions.
The downside is that you need larger sample sizes. Because different people are in each group, individual differences (personality, background, mood) add variability to your data. You need enough participants per group to detect effects despite that noise.
Every participant experiences all conditions, effectively serving as their own control group. This increases statistical power dramatically because individual differences are no longer a source of variability between conditions. The same person's response to condition A is compared directly to that same person's response to condition B.
The major threat is order effects. Fatigue, practice, boredom, or sensitization from earlier conditions can contaminate responses to later conditions. If participants always see the emotional message first and the rational message second, you can't tell whether differences are due to message type or simply to the order.
This is a specific type of within-subjects design where the same participants are measured multiple times across conditions or time points. It reduces error variance by using participants as their own baseline, making it easier to detect true treatment effects.
Counterbalancing is essential here. By varying the order of conditions across participants (half see condition A first, half see condition B first), researchers can control for sequence effects. Without counterbalancing, order effects become a serious confound.
Compare: Between-subjects vs. within-subjects: the trade-off is carryover effects vs. sample size. Between-subjects needs more participants but avoids contamination; within-subjects is more statistically efficient but requires careful counterbalancing. Know which threats each design faces.
Sometimes researchers need to examine how multiple factors work together. Factorial designs allow for more sophisticated questions about interaction effects.
A factorial design tests two or more independent variables simultaneously. A 2×2 design, for example, might cross message type (emotional vs. rational) with source credibility (high vs. low), creating four total conditions.
The real power of factorial designs is that they reveal interaction effects that single-variable designs would completely miss. Maybe emotional appeals are highly persuasive when the source is credible but backfire when the source lacks credibility. A study testing only message type or only source credibility would never uncover that pattern.
Factorial designs are also an efficient use of participants, since every subject provides data relevant to multiple hypotheses within a single study.
Compare: Factorial vs. single-factor designs: factorial designs answer more complex questions about how variables combine, but they require careful interpretation of main effects and interactions. If an FRQ describes a study examining how two message features work together, factorial design is the answer.
Where you conduct your experiment creates a fundamental trade-off between the precision of your measurements and the applicability of your findings to the real world.
Lab experiments are conducted in controlled environments where researchers can isolate variables and eliminate extraneous influences. They maximize internal validity through precise manipulation and measurement.
The limitation is ecological validity. Participants watching political ads in a sterile lab may respond very differently than when they encounter those same ads while scrolling through social media at home. The artificial setting can change behavior in ways that don't generalize to real life.
Field experiments take place in natural settings like workplaces, classrooms, online platforms, or public spaces. They enhance ecological validity because participants behave more naturally in familiar environments, and the treatment is embedded in a realistic context.
The sacrifice is control. Researchers cannot eliminate all extraneous variables or ensure perfectly consistent conditions across participants. Unexpected disruptions, varying environments, and uncontrolled social influences all introduce noise.
Compare: Laboratory vs. field experiments: this is the classic internal vs. external validity trade-off. Lab experiments tell you if something can work under ideal conditions; field experiments tell you whether it actually works in the real world. Strong research programs often use both approaches to build a complete picture.
| Concept | Best Examples |
|---|---|
| Random assignment (causal claims) | True experimental design |
| Pre-existing groups (real-world constraints) | Quasi-experimental design |
| Controlling for pre-test effects | Solomon four-group design, pre-test/post-test control group |
| Eliminating carryover effects | Between-subjects design |
| Maximizing statistical power | Within-subjects design, repeated measures design |
| Testing interaction effects | Factorial design |
| Maximizing internal validity | Laboratory experiments, true experimental design |
| Maximizing ecological validity | Field experiments |
A researcher wants to study how a new crisis communication training affects employee responses, but company policy requires using intact departments rather than randomly assigning individuals. Which design type is appropriate, and what validity threat should the researcher address?
Compare within-subjects and between-subjects designs: What specific threat does each design face, and how do researchers typically mitigate these threats?
A study finds that a persuasive message only changes attitudes when participants completed a pre-test questionnaire first. Which design would have detected this problem, and how does it work?
If an FRQ asks you to design a study examining how both message framing (gain vs. loss) and messenger identity (celebrity vs. expert) affect health behavior intentions, which design allows you to test these factors efficiently and examine their combined effects?
A communication researcher finds strong effects for a media literacy intervention in the lab but no effects when the same intervention is tested in actual classrooms. Explain this discrepancy using the concepts of internal and external validity.