Mastering Experimental Research Design: Types and Techniques

Every research project that claims “this caused that” rests on a quiet assumption: the researcher has ruled out every other explanation. In library and information science, this matters constantly. Does a new information literacy module actually improve students’ search skills, or would they have improved anyway? Does redesigning a library’s signage genuinely reduce reference desk queries? Answering these questions confidently requires more than observation. It requires a deliberate plan that isolates a cause and measures its effect. That plan is what we call experimental research design, and understanding its types and techniques is essential for anyone who wants their findings to hold up to scrutiny.

Understanding experimental research design 🔗

Experimental research design is the structured blueprint a researcher uses to test whether one variable produces a change in another. The variable the researcher deliberately changes is the independent variable, and the outcome being measured is the dependent variable. The defining feature is active intervention: the researcher manipulates the independent variable on purpose, rather than simply watching events unfold.

This is precisely what separates experimental work from descriptive or correlational studies. A survey might reveal that students who attend library workshops tend to write better-cited assignments. But correlation alone cannot tell us whether the workshop caused the improvement. Stronger students might simply be more likely to attend workshops in the first place. To move from “these things occur together” to “this produces that,” we need a design built specifically for causal claims.

Why establishing cause and effect needs a structured design 🔗

Three conditions must be satisfied before we can claim a causal relationship. First, the cause must come before the effect in time. Second, the cause and effect must actually vary together. Third, and most challenging, every alternative explanation must be eliminated. That third condition is where experimental design earns its keep. The randomized controlled trial is widely regarded as the most rigorous method for determining whether a cause-effect relationship exists between an intervention and an outcome, precisely because it is engineered to close off rival explanations.

Those rival explanations have a name in research methodology: confounding variables. These are outside factors that influence the outcome alongside, or instead of, the variable under study. A well-built experiment neutralises confounders so that any observed difference can be attributed to the treatment and nothing else.

The core building blocks 🔗

Most experimental designs rely on a small set of tools. The experimental group receives the treatment or intervention, while the control group does not, serving as a baseline for comparison. A pre-test measures participants before the intervention, and a post-test measures them afterwards. The most powerful tool of all is random assignment, the process of allocating participants to groups by chance. Random assignment matters because, as researchers note, it minimises population bias and increases the internal validity of the study by making the groups comparable on age, ability, motivation, and every other characteristic before the treatment is applied.

Types of experimental designs 🔗

Experimental designs are usually grouped into three broad families: pre-experimental, true experimental, and quasi-experimental. They differ mainly in how rigorously they control confounding variables. Within these families sit several specific designs that a researcher chooses based on resources, ethics, and the question at hand. Let us work through the most important ones.

The before-after design 🔗

The before-after design, also called the one-group pre-test post-test design, measures a single group before the treatment, applies the intervention, then measures the same group again. The difference between the two scores is treated as the effect of the treatment. A library might test the reference skills of a class, run a database training session, and then re-test the same class to see if scores rose.

This design is simple and inexpensive, which is why it appears so often in small studies. It is also a classic pre-experimental design, conducted as a first step to see whether an intervention shows promise before committing to a larger study. The catch is that it has no control group. If scores improve, we cannot be sure the training caused it. The students might have matured over the period, learned from other sources, or simply performed better the second time because they had seen the test before, a problem researchers call a testing threat to internal validity. The before-after design tells us something happened, but not reliably why.

The after-only design 🔗

The after-only design, or one-shot case study, is even simpler. A treatment is applied and the outcome is measured once, afterwards, with no pre-test and no control group. A librarian might introduce a new catalogue interface and then survey user satisfaction at the end of the month.

Because there is no baseline measurement and no comparison group, this design offers the weakest evidence of all. We have a single snapshot with nothing to compare it against. Did satisfaction rise, fall, or stay the same? Without a “before” reading or a control group, the question cannot be answered. The after-only design is best understood as exploratory. It can generate hypotheses and justify a more serious study, but it should not be the basis for firm causal conclusions.

The two-group design with a control 🔗

Adding a control group transforms the picture. In the pre-test post-test control group design, two groups are measured before the intervention; one receives the treatment and the other does not; then both are measured again. By comparing the change in the experimental group against the change in the control group, the researcher can separate the effect of the treatment from changes that would have happened anyway.

When participants are randomly assigned to these two groups, the design becomes a true experimental design, the gold standard for causal inference. The logic is elegant: because random allocation makes the groups equivalent at the start, the only significant difference between them is whether they received the treatment, so any difference in outcomes can be attributed to the treatment itself. A variant, the post-test only control group design, drops the pre-test entirely and relies on randomisation alone to ensure the groups start out equal, which sidesteps the risk that taking a pre-test influences later behaviour.

Quasi-experimental designs 🔗

In real library, education, and social settings, random assignment is often impossible. You usually cannot randomly assign students to classrooms or randomly decide which branch library gets a new service. This is where quasi-experimental designs come in. They resemble true experiments and still involve manipulating an independent variable and using a comparison group, but they lack the random assignment that defines a true experiment.

The most common form is the nonequivalent comparison group design, where the researcher uses intact, pre-existing groups, such as two different class sections, and assigns the treatment to one of them. Other widely used variants include the interrupted time series design, which takes many measurements before and after an intervention to detect a shift in the trend. Researchers often turn to these designs for practical or ethical reasons, since it would sometimes be unfair to withhold a beneficial service from people purely for the sake of a study. Quasi-experimental designs are increasingly employed to balance internal validity with real-world relevance, making them especially valuable in applied fields where laboratory-style control is unrealistic.

Internal and external validity: the techniques that matter 🔗

Choosing a design is ultimately about managing two kinds of validity. Internal validity is the confidence that the treatment, and not some other factor, caused the observed effect. External validity is the extent to which the findings generalise to other people, settings, and times. The unfortunate reality is that the two often pull in opposite directions. Tight laboratory control boosts internal validity but can make the setting so artificial that the results do not transfer to a working library.

Several techniques strengthen validity. Randomisation protects internal validity by balancing confounders. Control groups reveal what would have happened without the treatment. A clever extension is the Solomon four-group design, which combines pre-test post-test and post-test only groups to isolate any influence the pre-test itself might have had, thereby controlling for the testing effect while improving external validity. Matching participants on key characteristics can improve comparability when randomisation is not available.

Choosing the right design for your study 🔗

There is no single best design, only the best design for a given question and set of constraints. A doctoral researcher with funding and cooperative participants may run a true experiment. A practitioner evaluating a new service across two branches will more realistically use a quasi-experimental approach. A quick pilot to see whether an idea is worth pursuing might justify a simple before-after study. The skill lies in matching the rigour of the design to the strength of the claim you intend to make, and in being honest about the limitations that follow from your choice.

What do you think? If you were studying whether a new digital literacy programme improves student outcomes in your institution, which design would you choose, and what confounding variables would worry you the most? When is it acceptable to trade some internal validity for findings that better reflect the messy reality of everyday practice?

References

1. https://www.healthknowledge.org.uk/e-learning/epidemiology/practitioners/introduction-study-design-is-rct
2. https://study.com/learn/lesson/randomized-controlled-trial-overview-design-examples.html
3. https://en.wikipedia.org/wiki/Internal_validity
4. https://thedecisionlab.com/reference-guide/statistics/randomized-controlled-trial
5. https://www.scribbr.com/methodology/quasi-experimental-design/
6. https://pmc.ncbi.nlm.nih.gov/articles/PMC6923620/
7. https://en.wikipedia.org/wiki/External_validity
8. https://jihongzhang.org/teaching/2025-01-13-Experiment-Design/Lecture06/ESRM64103_Lecture06.html