Demand characteristics are the subtle cues in a study that tell participants what is being tested, prompting them to change their behaviour to fit what they think the researcher wants. When people guess the aim of an experiment, they often stop behaving naturally and instead play the role of a “good subject” — and that contaminates the results. This guide gives you a precise definition of demand characteristics, explains where the cues come from, walks through worked examples and famous cases, and shows the exact techniques used to reduce them in dissertations and experiments.
Have you ever sat in a class where a guest observer perches at the back with a clipboard? Suddenly everyone sits up a little straighter, and even the rowdiest students start raising their hands. The lesson did not get better — the participants did. That instinctive shift is a demand characteristic in action, and it is one of the most common threats to honest, trustworthy research.
It happens when people pick up on clues about how they are expected to act and — consciously or not — adjust their behaviour to be a “helpful” participant. Below is a snapshot of how a single cue can warp an entire finding.
| The cue (the “demand”) | The participant’s reaction | The biased result |
|---|---|---|
| A “stress” Questionnaire full of anxious wording | The participant starts fidgeting and acting nervous because the items imply they should feel stressed. | The study “proves” the environment is stressful, even when it was not. |
| A “brain-boosting” soda | The researcher says, “Drink this; it has special nutrients for focus.” The participant tries harder. | The participant appears more focused simply because they believe they have been boosted. |
| A “polite” Interview question | A tutor asks, “Did you find my lecture engaging today?” The student does not want to offend. | The student says “Yes!” to avoid being rude, regardless of the truth. |
What are demand characteristics?
A demand characteristic is any cue in a research setting that tips a participant off to the study’s aim or hypothesis and, in doing so, changes how they behave. The term was coined by the social psychologist Martin Orne in 1962, who argued that the experimental situation itself carries “demand” — an implicit pull towards a particular kind of behaviour. The participant is not lying; they are responding to a leak of information that suggests how a “good” person ought to act.
Crucially, demand characteristics are a form of research bias: they are a systematic distortion that pushes results in a predictable direction rather than a random error that cancels out across a sample. That is what makes them dangerous. If every participant nudges their behaviour the same way, the study can produce a confident, statistically significant — and completely false — conclusion.
“The totality of cues which convey an experimental hypothesis to the subject become significant determinants of the subject’s behaviour.” — Martin T. Orne, American Psychologist, 1962.
Think of it like being a guest at a formal dinner party. You might be starving and want to grab the bread rolls with both hands, but you notice the cloth napkins, the three different forks, and the host’s upright posture. Those environmental cues “demand” a certain level of etiquette. You are not being your natural self; you are being your dinner-party self. In a study, participants do exactly the same thing — they read the room and perform the version of themselves the situation seems to call for.
Demand characteristics vs the experimenter effect
These two ideas are often confused, so it is worth pinning down the difference. Demand characteristics come from the participant’s interpretation of the situation. The experimenter effect (also called experimenter bias or the Pygmalion effect when expectations lift performance) comes from the researcher unintentionally influencing results — through tone, body language, or how they record data. The two frequently feed each other: a researcher’s hopeful nod becomes the very cue a participant reads and responds to.
Worried about bias in your study design?
Our subject specialists help students design experiments and write up dissertations that stand up to examiner scrutiny.
What causes demand characteristics? Where the cues come from
Demand characteristics rarely come from a researcher saying, “Please act angry now.” They are far more subtle — woven into the environment, the people, and even the gossip around a study. Understanding the sources is the first step to designing them out.
1. Environmental cues
The setting speaks volumes. A high-tech laboratory with white walls and the smell of bleach pulls people towards reserved, “obedient” behaviour. A room filled with sports memorabilia might quietly nudge a participant towards competitive traits. The physical space frames what kind of behaviour seems appropriate before a single word is spoken.
2. Researcher interaction (the experimenter effect)
Human interaction is a minefield of non-verbal communication. A researcher might nod slightly when a participant gives an answer that supports the hypothesis, or look faintly disappointed when the data goes the other way. Even a small change in tone when reading instructions can act as a cue. None of this is deliberate, which is exactly why it is so hard to catch without procedural safeguards.
3. The task and the wording
If a participant is asked to watch a violent film clip and then play a game involving “aggression”, they can usually guess what is being measured. Loaded item wording on a questionnaire — words like “stressful”, “annoying”, or “exciting” — signals the expected answer just as clearly.
4. The “rumour mill”
In university settings, students often take part in the same studies for course credit. If word gets out that “the study in Room 302 is about testing aggression”, every participant who walks in afterwards is already contaminated by the rumour. This is a serious, under-appreciated threat in any departmental participant pool. The same dynamic applies in field research — for example, an environment-focused study where local communities have heard what the researcher is “really” looking into.
The four participant roles: from “good subject” to sabotage
When participants think they have worked out the goal of a study, they tend to fall into one of four roles. Each one biases the data in a different direction, which is why naming them matters when you write up your limitations section.
| Participant role | What they do | Effect on the data |
|---|---|---|
| The good subject | Wants to be helpful, so supplies the data they think the researcher is hoping for. | Inflates support for the hypothesis (the most common problem). |
| The negativistic subject | Resents being studied; deliberately acts the opposite way to “break” the experiment. | Suppresses or reverses the expected effect. |
| The faithful subject | Follows instructions meticulously and ignores any hunches about the aim. | The gold standard — minimal distortion. |
| The apprehensive subject | Worried about being judged; acts to look “normal” or “smart” rather than honestly. | Shifts responses towards social desirability. |
Notice that three of the four roles distort the data, and they do not all push the same way. That is why simply assuming participants will “cancel out” is naive — demand characteristics are a directional bias, not noise. The effect shows up across every discipline, from science lab experiments to social-survey work, wherever participants can sense what a study is hoping to find.
A worked example: the “revolutionary” learning app
The clearest way to see demand characteristics is to trace one through a realistic study from start to finish.
The cue: the words “amazing” and “completely transform”, plus the teacher’s visible excitement, signal the expected answer.
The reaction: the students, wanting to please a teacher they like, report that they loved the app and feel they learned more.
The biased result: the feedback form shows glowing scores, and she concludes the app “works”.
The problem: did the students learn more because of the app, or did they simply try harder because they knew the teacher cared about it? This overlaps with the Hawthorne Effect, where people perform better just because they are being observed and feel special. Either way, the apparent effect of the app is confounded, and the study cannot support the conclusion she has drawn.
The fix here is straightforward and shows how design choices remove cues: a neutral colleague (who does not know which app the school is championing) introduces the lesson with a scripted, flat description; the feedback form uses balanced wording; and learning is measured with an objective comprehension test rather than self-reported enjoyment. Suddenly there is no “right answer” for the students to perform.
Famous cases: when the cues took over
Some of the most cited studies in psychology are now re-examined precisely because of demand characteristics. They show how powerful these cues can be.
The Stanford Prison Experiment (1971)
Philip Zimbardo’s study is often presented as proof that situations turn ordinary people cruel. In recent years, critics — aided by released audio archives — have pointed to heavy demand characteristics. Several “guards” were given steering on how to behave, and many later said they were simply playing the part they believed Zimbardo wanted. On this reading the data was not revealing human nature; it was showing participants acting out a script they thought the “director” desired.
The Milgram obedience study (1963)
Stanley Milgram tested how far people would go in delivering apparent electric shocks when ordered by an authority figure. The results were striking, but some researchers argue a portion of participants saw through the set-up — they noticed the staged equipment or the experimenter’s lack of genuine alarm and suspected the shocks were not real. If they suspected the “game”, their willingness to continue was itself a demand characteristic: they were playing along with what the experiment seemed to require.
The lesson from both is the same. A finding is only as strong as the researcher’s ability to rule out the possibility that participants were performing rather than behaving. This is fundamentally a question of reliability and validity — specifically internal validity, the confidence that your independent variable, and not a hidden cue, produced the effect.
How to reduce and avoid demand characteristics
You can never fully delete human intuition, but a well-designed study removes most of the cues that participants act on. The table below summarises the standard toolkit; choose the techniques that fit your method and ethics approval.
| Technique | How it works | Best for |
|---|---|---|
| Single-blind design | The participant does not know the hypothesis or which condition (control or experimental) they are in. | Most experiments with manipulated conditions. |
| Double-blind design | Neither the participant nor the researcher interacting with them knows the condition, so the researcher cannot leak cues. | Drug trials and any study sensitive to the experimenter effect. |
| Deception with a cover story | Participants are told a plausible false aim (e.g. “this is about memory” when it studies conformity). Full debriefing is required afterwards. | Studies where knowing the true aim would obviously change behaviour. |
| Naturalistic / covert observation | People are observed in their everyday environment without knowing a study is taking place. | Behaviour that is hard to study honestly under lab conditions. |
| Standardised, neutral wording | Scripted or recorded instructions ensure every participant gets identical wording, tone, and emphasis. | Every study — the cheapest, highest-impact fix. |
| Filler items and distractor tasks | Padding the materials with unrelated items disguises which measures actually matter. | Surveys and questionnaires where the focus is easy to guess. |
Two design principles tie these together. First, minimise the participant’s ability to guess the aim — through blinding, neutral wording, and filler items. Second, minimise the cues the researcher gives off — through scripts, recorded instructions, and double-blind procedures. A study that does both, and that uses an objective outcome measure rather than self-report, is far more resistant to this bias.
It is also worth checking your own assumptions during recruitment and analysis: related biases such as affinity bias can creep in when a researcher unconsciously warms to participants who behave the way they expected. Keeping a clear, pre-registered protocol — deciding your measures and analysis before you collect a single response — is one of the strongest defences against all of these effects at once.
A simple checklist for your own study
- Write instructions as a script and read them identically to every participant (or record them).
- Strip loaded, leading words out of your questionnaire and interview items.
- Use a blind procedure so whoever runs the session does not know the condition.
- Add neutral filler items so the real focus is not obvious.
- Prefer objective outcome measures over self-reported enjoyment or confidence.
- Acknowledge any residual risk honestly in your limitations section.
The human element: why demand characteristics happen at all
It is tempting to treat demand characteristics as a flaw in people, but they actually reveal something rather hopeful: we are social, cooperative beings who want to be helpful and to understand what is being asked of us. That cooperativeness is wonderful in everyday life and a genuine nuisance in a controlled experiment. Good research design does not fight human nature — it simply removes the cues that human nature would otherwise respond to, so the behaviour you measure is the real thing rather than a polite performance.
If you are planning your own experiment or writing up your methodology, treating demand characteristics as a design constraint from day one — rather than a limitation you confess at the end — is what separates a convincing study from a shaky one.
Designing a bias-proof dissertation study?
Our dissertation experts help you build a watertight methodology — blinding, controls, and all — from proposal to final write-up.
For the wider picture of how this fits with confirmation bias, sampling bias, and the experimenter effect, see our hub guide on research bias, and tie it back to your aims with a well-framed hypothesis before you collect any data.
