An experiment compares conditions and infers cause from the results — but only if the only thing that differs between conditions is the IV. This is why operationalisation and control of extraneous variables matter so much.
Operationalising variables
To operationalise is to define a variable so precisely that another researcher could measure it in exactly the same way.
- Vague: "stress".
- Operationalised: "self-reported stress on a 1–10 Likert scale, completed immediately after the test."
- Vague: "intelligence".
- Operationalised: "score on the WAIS-IV Full Scale IQ test."
Without operationalisation, you cannot replicate the study or compare findings across studies.
Extraneous variables (EVs)
An extraneous variable is anything other than the IV that could affect the DV. Examples: room temperature, time of day, noise, participant tiredness, instructions phrasing, experimenter behaviour. Extraneous variables that vary unsystematically add noise to the data.
Confounding variables
A confounding variable is an extraneous variable that varies systematically with the IV. This is the worst case — you can no longer separate the effect of the IV from the effect of the confound. Example: testing music vs silence groups in different rooms (one warm, one cold). If recall differs, you can't tell whether music or temperature caused it.
Controls
Good control reduces extraneous variables to noise (or eliminates them):
- Standardised procedure: every participant gets the same instructions, equipment, environment, time of day.
- Single-blind / double-blind: participants (and experimenter) don't know the condition, eliminating expectancy effects and demand characteristics.
- Counterbalancing: in repeated-measures designs, vary the order of conditions across participants to balance practice/fatigue effects.
- Random allocation to conditions: balances participant variables (motivation, ability) across groups.
- Standardised materials: same word lists, same images, same software.
Why control matters
If you don't control EVs, your data is noisy: real effects can be hidden and you may report false ones. Control is the difference between an experiment and a casual observation.
⚠Common mistakes— Common errors
- Calling random allocation "random sampling" (different things).
- Listing controls but not explaining why each addresses a specific extraneous variable.
- Forgetting to operationalise the DV (most common single mark loss in 4-mark answers).
AI-generated · claude-opus-4-7 · v3-deep-psychology