Research

Find the confounds before you run the experiment

Branch your design into conditions, controls and confounds, and plan a fix for each problem before you collect any data.

Start from a template

An experiment-design template you can clone. Put in your research question, then work through conditions, confounds and the protocol before you collect data.

Clone this template

How it works

State what you want to test and the model proposes a design. Each condition, confound and protocol step gets its own branch, and every question under it keeps the design in context.

  1. State the hypothesis. Enter what you want to test. Ask the model to name the independent and dependent variables and the main choices: within or between subjects, how many conditions, and how you will measure the outcome.

  2. Give each condition a branch. Under each one, set out the exact manipulation, what it is compared against, and how you will check that the manipulation worked.

  3. List the confounds. For each comparison, ask what else could explain a difference, such as order effects or experimenter bias. Write a mitigation under each confound.

  4. Lay out the protocol. Add a branch for what a participant goes through, from consent and baseline measures to the manipulation check and debrief. Read it top to bottom to see whether the session runs too long or a measure is missing.

Why a map instead of a protocol document

A control only makes sense against a specific comparison. With each condition on its own branch, you can see every comparison the design supports and decide which ones answer your question.

Fixing one confound often changes something else, such as the counterbalancing scheme or the exclusion criteria. Ask about the change under the confound’s branch, and the knock-on effects sit right beneath it.

Example

Say you are testing whether nudges raise hand-washing rates on hospital wards. You branch into three conditions: a control ward, a ward with signs, and a ward with signs plus feedback. Under confounds, you note that wards with sicker patients could differ in compliance whatever the nudge. A branch on a crossover design, where wards rotate through the conditions, deals with that, and a branch under it plans for the carryover effect the crossover introduces.

Now try it yourself

Design an experiment to test whether background music affects reading comprehension in university students
A within-subjects design with three conditions: silence, instrumental music, and lyrical music. Key decisions involve counterbalancing order effects, selecting standardized reading passages of equal difficulty, and choosing a validated comprehension measure.
What confounding variables need to be controlled?
Musical training and listening habits are individual differences that could moderate the effect. Time of day affects alertness and baseline performance. Passage topic familiarity introduces another confound — use pretesting to match passage difficulty across conditions.

Ready to try experiment design?