When we conduct experiments, there are other variables that can affect our results if we do not control them.
An extraneous variable is any factor other than the independent variable (IV, the one the researcher deliberately changes) that could affect the results.
It can be a participant characteristic, such as intelligence, gender, or age, or a feature of the environment, such as lighting or noise. Left uncontrolled, it can obscure a real effect.
Key Takeaways
- Definition: Any variable other than the independent variable (IV) that could affect the dependent variable (DV), such as noise, lighting, or a participant’s mood.
- Extraneous vs Confounding: An extraneous variable only becomes a problem once it varies systematically with the IV. At that point it becomes a confounding variable, offering a rival explanation for the results.
- Participant Variables: Individual differences such as age, mood, or intelligence can differ between participants and are mainly controlled through random allocation.
- Situational Variables: Features of the environment or procedure, such as noise or instruction wording, are controlled by standardization (identical procedures) and counterbalancing (varying condition order).
- Investigator Effects: A researcher’s own expectations can unconsciously bias how they treat participants or interpret results. Double-blind procedures, where neither party knows the condition, help prevent this.
- Demand Characteristics: Participants often pick up cues about a study’s purpose and change their behavior to match or defy it. Single-blind procedures reduce this risk.
- Control Techniques: Standardization, counterbalancing, random allocation, and blinding each target a specific threat, so researchers match the technique to the type of variable.
Why Control Extraneous Variables
Researchers need to be sure that the independent variable alone is responsible for any change in the dependent variable. So every other variable that could affect the dependent variable must be controlled. Those are the extraneous variables.
Left uncontrolled, they threaten a study’s internal validity, our confidence that the IV caused any change in the DV.
The effect depends on how the variable behaves. If it varies randomly across conditions, it adds noise that makes a real effect harder to detect. If it varies systematically with the IV, it becomes a confounding variable, offering a rival explanation for the results.
Types and Examples of Extraneous Variables
1. Situational Variables
Situational variables are factors, conditions, or characteristics in the external environment that can influence behavior, decision-making, or outcomes. They are called “situational” because they belong to the context, not to a stable personal trait like personality.
Examples include physical aspects of the environment, such as weather, location, time of day, or noise level. Social aspects count too.
These include the presence of others, group dynamics, or societal norms, as well as more abstract factors like time pressure, level of risk, or unclear instructions. Situational variables should be controlled so they are the same for all participants.
Standardized procedures keep conditions the same for every participant. This means using the same instructions, timing, and testing environment for everyone.
2. Participant Variables
Participant variables are the individual differences between participants that could affect the results, such as mood, intelligence, anxiety, nerves, or concentration. They matter most in independent-groups designs.
For example, a tired or dyslexic participant with poor eyesight might perform worse on a memory test, regardless of the independent variable. The experimental design chosen can affect how much these differences matter.
Repeated-measures designs raise a related problem. This is called an order effect: practice can improve performance, or fatigue can worsen it, depending on the order conditions are completed in.
Counterbalancing controls this by varying that order, such as giving half the participants condition “A” first and the other half condition “B” first.
Participant variables are mainly controlled using random allocation to the conditions of the independent variable.
3. Experimenter / Investigator Effects
The experimenter can unconsciously convey how participants should behave, a bias known as experimenter bias. Subtle, unintentional clues about the study’s purpose can shape how participants respond, even when the experimenter has no idea they are giving them.
Rosenthal and Fode (1963) tested this directly. Students who were told they were running “maze-bright” rats got better results than students told they had “maze-dull” rats. In fact, the rats had been assigned to each group at random.
The rats hadn’t changed. What changed was how the experimenters handled and scored them, shaped entirely by their own expectations.
The experimenter’s own personal attributes, such as age, gender, accent, or manner, can also affect how participants behave.
People show the same effect. In the Pygmalion study, Rosenthal and Jacobson (1968) found that teachers’ induced expectations raised the measured IQ of randomly chosen pupils labeled “bloomers.”
4. Demand Characteristics
Demand characteristics are the clues in an experiment that reveal its purpose to participants. Participants rarely stay passive.
Orne (1962) argued that some try to guess what the researcher wants and give it to them, a pattern he called the “good subject” effect. Others deliberately do the opposite.
Many act from “evaluation apprehension,” behaving as they think they should be seen to behave, rather than as the independent variable dictates. Either way, participants are reacting to the situation, not to the independent variable. That reaction can look like a real effect.
Participants respond to their surroundings, the researcher’s characteristics and behavior (such as non-verbal communication), and their own interpretation of what is going on. Good controls limit this.
Experimenters can minimize these factors by keeping the environment as natural as possible and carefully following standardized procedures. Using different experimenters to see if they obtain similar results is another safeguard.
Example: Alcohol and Driving Ability
Suppose we wanted to measure the effects of alcohol (IV) on driving ability (DV). We would have to ensure that extraneous variables did not affect the results. These variables could include the following:
- Familiarity with the car: Some people may drive better because they have driven this make of car before.
- Familiarity with the test: Some people may do better than others because they know what to expect on the test.
- Used to drinking: The effects of alcohol on some people may be less than on others because they are used to drinking.
- Full stomach: The effect of alcohol on some subjects may be less than on others because they have just had a big meal.
If these extraneous variables go uncontrolled, they may become confounding variables and affect the results.
How to Control Extraneous Variables
Experimental design is largely a set of techniques for stopping extraneous variables from becoming confounding variables. Each technique targets a specific threat.
| Threat | Example | Main control |
|---|---|---|
| Situational variables | Noise, lighting, time of day, wording of instructions | Standardization, controlling the environment |
| Order effects | Practice or fatigue in repeated-measures designs | Counterbalancing |
| Participant variables | Age, mood, intelligence | Random allocation, matching |
| Demand characteristics | Participants guessing the study’s purpose | Single-blind procedure, placebo |
| Investigator effects | The researcher’s expectations | Double-blind procedure |
Controlling Situational Variables
Standardization keeps the procedure, instructions, and materials identical for every participant. Situational variables then cannot differ systematically between conditions.
Researchers also hold the physical environment constant, including noise, lighting, temperature, and time of day. These are control variables, held constant on purpose. Testing everyone in the same quiet room is one example.
Time of day can be controlled by testing all participants at the same time.
Alternatively, test half of each group in the morning and half in the afternoon, so time of day cannot favor either group.
Order effects need a different fix. In a repeated-measures design, counterbalancing varies the order of conditions across participants, for example in an ABBA sequence. Practice and fatigue then cancel out instead of favoring one condition.
Controlling Participant Variables
Random allocation assigns participants to conditions by chance, so the groups do not differ systematically before the IV is applied. Fisher (1935) established it as the device that turns unknown extraneous variables into random error rather than bias. That is what makes causal inference possible.
It works even for participant variables nobody thought to measure.
In a matched-pairs design, researchers instead match participants on a key variable, such as age or IQ, so the groups start out equal on it.
Some participant variables cannot be controlled by design. A researcher can instead measure the variable as a covariate and control it statistically, for example with analysis of covariance. This separates its influence on the DV from the effect of the IV, rather than leaving it as noise.
Blinding and Placebo Controls
Blinding targets expectation-based confounds. In a single-blind procedure, participants do not know which condition they are in, such as treatment or control.
That reduces demand characteristics. These cues reveal a study’s purpose. Orne (1962) argued they lead participants to confirm, oppose, or act out what they think the researcher wants.
A double-blind procedure keeps the experimenter who interacts with participants unaware too. Without it, a researcher’s expectations can quietly shape how participants are treated and how behavior is scored, as in the rat study above.
Placebos add another safeguard. A placebo condition mimics the real treatment without its active ingredient. In a caffeine experiment, for instance, the placebo group might drink decaffeinated coffee. Researchers often combine it with blinding, so participants cannot tell which treatment they received.
References
Fisher, R. A. (1935). The design of experiments. Oliver & Boyd.
Orne, M. T. (1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 17(11), 776–783.
Rosenthal, R., & Fode, K. L. (1963). The effect of experimenter bias on the performance of the albino rat. Behavioral Science, 8(3), 183–189.
Rosenthal, R., & Jacobson, L. (1968). Pygmalion in the classroom: Teacher expectation and pupils’ intellectual development. Holt, Rinehart & Winston.
