Reading experiments
Six questions to ask of any study before deciding what it shows.
Variables
The independent variable is what the researchers change or compare, such as the verb in a question or the size of a payment. The dependent variable is what they measure, such as a speed estimate or a rating. Ask how the measure was defined, because a vague measure leaves room for several readings of the same result.
Controls
A control holds something constant or provides a baseline. Bandura's children who saw no adult are a control group. Stroop's colored squares are a baseline task. When a study has no control, as with Little Albert, you cannot tell what would have happened without the manipulation.
Comparison
Results mean something only relative to something else. Compare between groups, or within the same people across conditions. Check that the groups were alike before the manipulation. Random assignment is the usual way to make them alike on average, and some profiles on this site say it was not described or not used.
Causality
A causal claim needs a manipulated difference, a fair comparison and few rival explanations. Experimenters' expectations, instructions and participants' guesses about the purpose can all produce a result. That is the central problem for the Stanford prison simulation. Observational follow-ups, such as the 2018 look at waiting in childhood, can show association but not cause.
Designs differ in how much they can support:
- Experiment
- The researchers set the conditions and compare outcomes between or within groups. Each profile says whether participants were randomly assigned.
- Quasi-experiment
- The researchers compare conditions, but people or animals were placed in them by non-random, pre-existing or external assignment, so groups may differ in other ways.
- Demonstration
- A single case or single condition that shows a phenomenon can occur. It has no comparison that rules out other explanations.
- Paradigm
- A general procedure used across many studies, laboratories and species, not one dated trial.
- Simulation
- A staged situation meant to resemble a real setting. What it shows about real settings depends on how closely the staging matches them.
Replication
A finding is stronger when other teams, with other samples, get a similar result. A replication can match the original, shrink it, or fail to find it. Partial replications, like Burger's version of Milgram's study that stopped at an earlier shock level, test only part of the original. Read the reassessment and source notes to see which later claims were checked and how closely the later procedure matched the original.
Ethics
Several studies involved substantial distress or vulnerable participants. Look for what participants were told, whether they could stop, what harm was possible, who was vulnerable (infants, children, animals) and what was done afterward. Assess consent, independent review and protection from harm alongside the reported findings. Learning from the findings and judging the methods are separate tasks.