1.2 - Variables and operationalisation

What changes and what is measured

An experiment investigates whether changing one variable affects another. The independent variable (IV) is the variable changed between conditions. The dependent variable (DV) is what is measured to investigate the effect.

Suppose a researcher asks whether background speech affects recall. Participants study a word list either in silence or with a recording of speech, then write down remembered words. Background sound is the IV; recall performance is the DV. The two sound arrangements are the levels or conditions of the IV. The DV is not the participant, the word list or simply the name of the experiment.

In a controlled experiment the researcher manipulates the IV. Some investigations compare a naturally occurring difference rather than create it; the researcher cannot randomly assign a person's age or past experience. In a correlation, two measured variables are co-variables: there is no manipulated IV and no demonstrated causal direction. Calling the larger numerical measurement the IV does not make a correlation an experiment.

Turning variables into procedures

An operational definition specifies exactly how a variable will be manipulated or measured. A name such as memory, anxiety or concentration identifies an idea but does not tell another researcher what to do.

For the sound investigation, an operationalised IV could be studying 20 words for 60 seconds either in a quiet room with no recording or with the same neutral speech recording played at the same set volume. An operationalised DV could be the number of those 20 words correctly written during a 90-second free-recall period immediately afterwards. A scoring rule should say whether spelling errors are accepted when the word is unambiguous, and whether repeated words count only once.

This makes the comparison testable and supports replication: another researcher can reproduce the materials, timing and scoring. It also makes numerical results interpretable. A score of 12 means 12 correctly recalled target words, not an unspecified amount of memory.

Precision does not guarantee construct validity, whether the measure represents the intended psychological idea. Counting recalled words measures performance on that memory task, but may not capture memory for real conversations. Counting pen taps is a clear procedure but could reflect habit rather than anxiety. Choose a measure for a defensible reason, pilot it for comprehension and practicality, and state what the measure cannot establish. An operationalised IV also needs meaningful levels: a vague instruction to play sound loudly will be hard to reproduce.

Extraneous variables and confounds

An extraneous variable is any variable other than the IV that could influence the DV. Participant factors include prior ability, tiredness and motivation; situational factors include room temperature or interruptions; procedural factors include instructions, task length and researcher behaviour.

An extraneous influence may add random variation if it varies without a consistent relationship to condition. A confounding variable systematically varies with the IV and offers an alternative explanation of the result. If the silence group studies for 60 seconds but the speech group studies for 30 seconds, study time is confounded with sound. If silence produces better recall, the researcher cannot separate the effect of sound from the effect of extra study time.

This distinction matters. An unexpected noise affecting one randomly chosen participant may make the data less consistent, but it is not automatically a systematic confound. If one whole condition is tested beside noisy building work and the other is not, the noise is systematically associated with condition. Internal validity is weakened because the intended IV is not the only plausible cause of the difference.

Controls that preserve the intended comparison

A control should limit an unwanted influence while preserving the intended IV difference. Give both sound conditions the same study and recall time, word difficulty, instructions and scoring rules. Use the same speech recording at a specified volume whenever that condition is used; do not remove sound from both groups, because that would remove the manipulation.

Standardisation keeps the relevant procedure consistent. Written instructions or a recorded script reduce differences in what the researcher says. An objective scoring key reduces judgement variation. Random allocation reduces systematic participant imbalance between independent groups; matching can reduce differences on selected relevant characteristics; repeated measures controls stable individual differences within a person. For repeated measures, counterbalancing spreads condition order so practice or fatigue is not tied to just one condition. These procedures have limits: random allocation does not guarantee equal ability, and counterbalancing cannot erase irreversible learning.

Researcher expectations can also influence treatment or scoring. Where feasible, a scorer who does not know the participant's condition can apply the same key without expecting one group to perform better. This is not always practical; a visible manipulation may reveal condition. Avoid claiming that a blind scorer makes the entire study double-blind.

Tighter controls can improve causal interpretation and consistency, but an unusual, interruption-free word task may differ from everyday studying. Explain that trade-off rather than calling all controls a weakness. The appropriate response is to control relevant alternatives, report the exact procedure and restrict the conclusion to what the design supports.