Methodology
Demand Characteristics
When participants guess the experimenter's hypothesis and act accordingly, threatening internal validity
Demand characteristics are cues in an experimental setting that lead participants to infer the researcher's hypothesis and adjust their behavior to match — or sometimes to subvert — it. Martin Orne (1962) defined the term in a landmark American Psychologist paper, arguing the experiment is a social situation and the participant a cooperative role-player rather than a neutral subject. Orne demonstrated that participants asked to add columns of random digits and then tear up the result complied for hours rather than refuse, simply because the experiment frame demanded it. The phenomenon threatens internal validity by confounding the manipulation with participants' guesses. Standard remedies include cover stories, deception, between-subjects designs, placebo controls, and post-experiment suspicion checks.
- Coined byMartin Orne (1962)
- Origin paperOn the social psychology of the psychological experiment
- Classic demonstrationEndless tedious task compliance
- "Good-subject" effectMost try to confirm the perceived hypothesis
- Major remedyCover stories, blind designs, placebos
- Suspicion checkPost-debrief funnel questionnaire
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why demand characteristics matter
- Experimental design. Any laboratory study must consider whether participants can infer and confirm the hypothesis.
- Clinical trials. Without double-blinding, expectancy effects inflate apparent treatment efficacy substantially — sometimes 30-50%.
- Survey research. Question wording and order cue desired answers; social-desirability response sets parallel demand effects.
- Replication crisis. Studies relying heavily on participant cooperation with subtle manipulations replicate poorly across labs.
- Education research. Hawthorne effects make any new program look effective during initial enthusiasm.
- UX testing. Users tell observers what they think researchers want to hear; behavioral logs reveal different preferences.
- Forensic interviewing. Suggestive questions create false memories in eyewitnesses and especially children, paralleling demand dynamics.
Common misconceptions
- Only naive participants are affected. Educated and experienced participants guess hypotheses faster, sometimes producing larger demand effects.
- Anonymous online studies eliminate it. Mechanical Turk and Prolific samples include experienced participants who recognize common paradigms within seconds.
- Deception always solves it. Suspicious participants who do not believe the cover story may show even larger demand-driven behavior.
- It is the same as experimenter bias. Experimenter bias is the researcher's behavior leaking expectations; demand characteristics include impersonal cues like setting and instructions.
- It only inflates effects. Negativistic participants intentionally subvert hypotheses, suppressing or reversing real effects.
- One funnel debrief is enough. Suspicion can be sophisticated; multiple converging checks (manipulation checks plus debrief) catch more contamination.
Frequently asked questions
What are demand characteristics?
Any features of an experimental setup — instructions, rumors, lab decor, subtle experimenter behavior — that hint at the researcher's hypothesis and shape participant behavior. Orne (1962) noted participants enter experiments wanting to be helpful and to "act like good subjects." When they can guess what the researcher expects, their behavior is contaminated by that guess rather than reflecting only the manipulation.
What did Orne demonstrate?
In a striking series, Orne asked participants to perform pointless tasks — adding 224 columns of random digits, then tearing each sheet into precisely 32 pieces. He expected refusal within minutes; participants continued for over five hours. The frame "this is a psychology experiment" supplied enough demand to override boredom. Orne concluded experimental compliance is itself a powerful behavior worth studying.
What is the good-subject effect?
The tendency of participants to do what they believe the researcher wants. Weber and Cook (1972) distinguished four roles — good subject (confirms hypothesis), faithful subject (follows instructions literally), apprehensive subject (tries to look mentally healthy), and negativistic subject (subverts the study). The good-subject role is most common in cooperative samples, biasing results toward predicted effects.
How does it relate to placebo effects?
Closely. A pill described as a powerful painkiller produces real analgesia partly because the participant infers they should feel better. Double-blind, placebo-controlled trials minimize demand by keeping both participants and assessors ignorant of group assignment. The Hawthorne effect — workers improving simply because they are observed — is another demand-driven phenomenon documented in the 1920s factory studies.
How do researchers reduce it?
Cover stories that disguise the true hypothesis; between-subjects designs so participants cannot compare conditions; deception (with debrief); placebo and active control conditions; double-blind administration; unobtrusive measures and field studies; physiological or implicit measures less under conscious control; and post-experiment funnel debriefs that probe what participants thought the study was about.
What is a funnel debrief?
A graded post-study questionnaire moving from open ("What do you think this study was about?") to specific ("Did you notice anything connecting the first task and the second?") to direct ("Did you suspect the confederate was an actor?"). Bargh and Chartrand (2000) recommend it for any priming or deception study. Participants who guessed correctly are excluded or analyzed separately, protecting the inference.
Are demand characteristics always bad?
Not in applied research. In therapy, classroom interventions, or workplace programs, the goal is real-world outcomes, and participants knowing the goal may be part of the intervention. The danger is in basic-research causal inference, where any effect attributable to guessing the hypothesis rather than to the manipulation undermines the theoretical claim.