Science for Everybody

Contents

• Introduction
• Why Control of Variables Matters
• Extraneous Variables
• Confounding Variables
• Demand Characteristics
• Investigator Effects
• Randomisation
• Standardisation
• Using Control Strategies Together
• Quick Check Task
• Extended Reading
• Evaluation
• Strengths
• Weaknesses
• Summary
• Quick Questions
• Key Terms

---

Introduction

Psychological research aims to isolate relationships between variables, but unwanted influences can distort findings. Controlling extraneous variables, preventing confounds, and reducing participant and researcher effects helps psychologists make more valid conclusions.

Randomisation and standardisation are two key methods used to improve control, while careful design can reduce demand characteristics and investigator effects.

This topic aligns with AQA A-Level Psychology Research Methods content on manipulation and control of variables, randomisation, standardisation, demand characteristics and investigator effects.

Why Control of Variables Matters

In an experiment, the researcher changes the independent variable and measures the dependent variable. Ideally, the IV should be the only systematic difference between conditions. In practice, many other factors can affect the DV. These are extraneous variables.

An extraneous variable does not automatically invalidate a study. If it varies randomly, it may add noise and make an effect harder to detect. However, if it varies systematically with the IV, it becomes a confounding variable and can offer an alternative explanation for the result. This directly threatens internal validity.

For example, imagine a memory experiment in which the caffeine condition is always tested in the morning and the no-caffeine condition is always tested in the afternoon. Time of day is no longer just an extraneous variable because it changes systematically with the condition. The researcher cannot tell whether caffeine or time of day caused any difference in memory performance.

Extraneous Variables

Extraneous variables are unwanted variables that may influence the DV. Researchers try to identify them during planning and either control them or reduce their effect. Two useful categories are participant variables and situational variables.

Participant variables are characteristics of the people taking part. Examples include age, intelligence, motivation, personality, prior experience, tiredness and baseline anxiety. These variables can affect performance even when the IV has no effect.

Situational variables are features of the research environment or procedure. Examples include noise, temperature, lighting, time of day, room layout, equipment and the exact instructions given to participants.

Not every extraneous variable needs identical control. The key question is whether it could influence the DV and whether it is likely to differ between conditions. Researchers should focus on variables that could plausibly distort the relationship between the IV and DV.

Confounding Variables

A confounding variable is a special type of extraneous variable. It changes systematically with the IV, so the researcher cannot separate its effect from the effect of the IV. This makes causal conclusions much weaker.

The source example uses an energy-drink study. If the first condition happened to contain mostly shy participants and the second condition mostly extroverted participants, personality could influence how much people talk. If personality differs systematically between the two conditions, it becomes a confounding variable.

This distinction matters: all confounding variables are extraneous variables, but not all extraneous variables are confounding variables. A random background noise that affects both conditions unpredictably is extraneous; a different noise level that always occurs in one condition is potentially confounding.

Demand Characteristics

Participants are active within research. They may notice cues in the procedure, materials or researcher behaviour and work out what the study is investigating. These cues are demand characteristics.

Once participants think they know the aim, they may change their behaviour. Some may try to help the researcher by behaving as they think is expected, sometimes called a please-u effect. Others may deliberately do the opposite or disrupt the expected pattern, sometimes called a screw-u effect. Either response means the DV may reflect participants’ guesses rather than the genuine effect of the IV.

Researchers can reduce demand characteristics by keeping the full aim hidden until debriefing when ethically justified, using standardised instructions, including plausible filler tasks, or using measures that are less obvious to participants. However, any deception must be justified and followed by an appropriate debrief.

Investigator Effects

Investigator effects occur when a researcher influences the outcome of an investigation. The influence may be conscious or unconscious. It can happen through tone of voice, facial expression, body language, wording of instructions, feedback, choice of participants, construction of materials or interpretation of ambiguous data.

Expectancy effects are one example. If a researcher expects one condition to perform better, subtle differences in encouragement or behaviour may influence participants. The textbook illustrates this with a maths-test scenario in which the same scripts could be judged differently if the marker believes one group should be more capable.

Standardised scripts, researcher training, objective scoring rules, automated procedures and blind or double-blind procedures can reduce investigator effects. These controls are especially important when researchers know the hypothesis and interact directly with participants.

Randomisation

Randomisation means using chance to remove researcher choice from parts of the procedure. It is a control technique because it reduces systematic bias and helps distribute uncontrolled influences more evenly.

For example, words in a memory list can be placed in a random order rather than an order chosen by the researcher. In a study with several conditions, the order in which participants complete conditions can also be randomised when appropriate. Random allocation can be used to place participants into independent-groups conditions.

Randomisation does not guarantee perfect equality. In a small sample, chance can still produce imbalanced groups. However, it reduces the risk that the researcher’s preferences or a predictable order will systematically affect one condition.

Standardisation

Standardisation means keeping the participant experience the same wherever it should be the same. Researchers use standardised instructions, identical timing, consistent equipment, the same room conditions and clear scoring procedures.

Standardisation helps control situational variables and investigator effects. It also improves replicability because another researcher can follow the same procedure. The only planned difference between experimental conditions should normally be the manipulation of the IV.

Standardisation does not mean every detail must always be identical. In repeated measures, for example, equivalent materials may be needed to avoid order effects. What matters is that any differences are planned, justified and controlled rather than accidental.

Using Control Strategies Together

Strong research usually combines several controls. Researchers identify likely participant and situational variables, standardise procedures, use randomisation where researcher choice could create bias, and consider whether participants or investigators could infer the expected outcome.

A pilot study can also reveal unexpected extraneous variables before the main investigation begins. Controls should be designed around the specific threat to validity rather than added mechanically.

Reflection question: imagine one condition in a stress study is tested in a quiet room and the other in a room beside a noisy corridor. What alternative explanation could the noise create, and how could the procedure be redesigned so the IV is the only systematic difference?

Quick Check Task

For each scenario, identify the likely extraneous or confounding variable and suggest one realistic control.

• A memory study tests Condition A at 9:00 a.m. and Condition B at 3:00 p.m.
• One researcher gives warmer encouragement to participants in the condition they expect to perform better.
• Participants are told the study is about whether caffeine improves attention before completing the task.
• A researcher chooses the order of words in a memory list based on which sequence seems easiest to remember.

Suggested Answers

• Time of day is confounding because it varies systematically with condition; test both conditions across equivalent times or randomise/counterbalance timing.
• This is an investigator effect; use a standardised script, blind procedure or automated instructions.
• This may create demand characteristics; conceal the full aim when ethically justified and debrief afterwards.
• Researcher choice could create order bias; randomise the word order.

---

Evaluation

Strengths

Improves internal validity

A major strength of controlling variables is that it improves internal validity. If researchers keep relevant extraneous variables constant or distribute them fairly, alternative explanations become less plausible. Therefore, any difference in the DV is more likely to result from the IV. This makes causal conclusions more convincing, especially in laboratory experiments.

Supports replication and reliability

Standardisation creates clear, repeatable procedures. When every participant receives the same instructions, timing and treatment, other researchers can reproduce the method more accurately. As a result, findings can be checked for consistency, which strengthens psychology as a scientific discipline.

Randomisation reduces systematic researcher bias

Randomisation removes some researcher decisions from the design. For example, randomly ordering stimuli prevents the researcher from unintentionally arranging material in a way that favours a prediction. This reduces systematic bias and can make the comparison between conditions fairer.

Controls can be matched to the specific threat

Psychologists are not limited to one control method. Standardisation can reduce situational variation, randomisation can reduce order or allocation bias, and blind procedures can reduce expectancy effects. Therefore, a well-designed study can combine controls to target several threats to validity at the same time.

Weaknesses

Too much control can reduce ecological validity

Strict control can make research artificial. Highly standardised instructions, unfamiliar laboratory settings and tightly controlled tasks may cause participants to behave differently from everyday life. Therefore, internal validity can improve while ecological validity and mundane realism decrease.

Residual confounding can remain

Researchers cannot identify or control every possible extraneous variable. Participant mood, previous experiences, subtle environmental changes or unknown third variables may still affect the DV. Randomisation also works probabilistically, so small samples can remain unbalanced by chance. As a result, control reduces confounding but cannot guarantee that it has been eliminated.

Demand characteristics and control can interact

Attempts to control a study may make its purpose more obvious. Repeated instructions, unusual tasks or very artificial procedures can encourage participants to search for the hypothesis. Therefore, tighter standardisation may sometimes increase participant reactivity rather than reduce it.

Ethical limits can restrict control

Researchers may hide the full aim to reduce demand characteristics, but this can involve deception or incomplete informed consent. Ethical guidelines require researchers to justify such procedures, protect participants and debrief them appropriately. Therefore, the most methodologically powerful control is not always ethically acceptable.

Control does not solve sample, cultural or temporal bias

A tightly controlled study can still use an unrepresentative sample, culturally specific materials or procedures that become outdated. Standardised instructions may not be interpreted in the same way across cultures, and findings can lose temporal validity as social behaviour changes. Therefore, strong experimental control does not automatically make findings generalisable across people, cultures or time.

Quick Questions

What is an extraneous variable?

Any variable other than the IV that may affect the DV if it is not controlled.

When does an extraneous variable become a confounding variable?

When it changes systematically with the IV and provides an alternative explanation for differences in the DV.

What are demand characteristics?

Cues that allow participants to guess the aim or expected behaviour and then alter their behaviour.

How does randomisation help control a study?

It uses chance to reduce researcher bias and systematic effects when allocating participants, ordering conditions or arranging materials.

How does standardisation improve research?

It keeps relevant procedures and instructions consistent, reducing extraneous variation and making replication easier.

Key Terms

Independent variable (IV) – The variable that the researcher manipulates or changes in an experiment.

Dependent variable (DV) – The variable that the researcher measures to see whether it changes as a result of the IV.

Extraneous variable (EV) – Any variable other than the IV that may affect the DV if it is not controlled.

Confounding variable – An extraneous variable that changes systematically with the IV, creating an alternative explanation for the findings.

Participant variable – An individual difference between participants, such as age, intelligence, personality or motivation, that may affect the DV.

Situational variable – A feature of the research setting, such as noise, temperature, time of day or instructions, that may affect the DV.

Demand characteristics – Cues in a study that allow participants to guess the aim or expected behaviour and then alter their behaviour.

Investigator effects – Any influence of the researcher’s behaviour or expectations on participants, procedure, data collection or interpretation.

Randomisation – The use of chance to reduce researcher bias when designing materials, deciding order, or allocating aspects of a procedure.

Standardisation – Using the same formalised procedures and instructions for all participants, apart from the planned difference in the IV.