Experimental and research designs

The classic command-word trap. Describe the difference wants a single comparative sentence, not two definitions side by side. Two separate definitions count as an implicit comparison, scoring 1 of 2. A connective earns the second mark: different participants are used in each condition, whereas the same are used in every condition.

Explain a strength or weakness is not describe one. Naming a feature earns little; marks come from why it matters here. Weak: a weakness of repeated measures is order effects. Full: …because participants do the task twice, so practice could raise the score, unrelated to the IV. Give the feature, what it looks like in the study, and the consequence.

Improvement suggestions are penalised when pre-learned rather than chosen for the study. A bigger sample is not automatically an improvement — it fixes neither order effects nor unmatched groups; it is the nature of the sample, not its size, that decides generalisation. The change must be within the researcher's control and target this design's weakness.

Because matched pairs uses different people in each condition, candidates often call it independent measures. The give-away is the word matched: each participant is deliberately paired with a similar person on what matters, such as age or ability. Ordinary independent measures has no such pairing. If participants are paired before allocation, it is matched pairs.

Asked to state the experimental design, candidates sometimes give the IV or DV (what was changed or measured), not how participants were allocated. Design is only allocation: same people in every condition (repeated), different people in each (independent), or different-but-paired (matched pairs). Answer with a design name, from who is in each condition.

Counterbalancing controls order effects, which arise only in repeated measures, where each participant does more than one condition. Suggesting it for independent measures or matched pairs is wasted: no participant there does more than one condition, so there is no order to control. Use counterbalancing only when the same people complete every condition.

Because repeated measures removes the problem of different-people groups, students conclude it is best everywhere. It is not. It creates order effects the other designs avoid — doing the task twice brings practice or tiredness — and where doing a task once spoils a second attempt it is wrong. It also makes demand characteristics more likely.

Both designs reduce participant variables, so students blur them, treating matched pairs as if the same people were in both. But matched pairs uses different people, each matched to a partner on age or ability. So it has no order effects, unlike repeated measures, but never perfectly. Same person in both → repeated; matched partners split → matched pairs.