Demand characteristics
Participants guess the study's aim and act to fit it.
What it means
Demand characteristics are cues within a study that allow participants to infer its purpose or the hypothesis, leading them to alter their behavior in response to that inference rather than to the manipulation itself. The mechanism is that participants are active interpreters of the research situation, not passive responders: having formed a guess about what is expected, they may behave in ways that confirm the hypothesis (to be 'good subjects' or please the researcher), that present themselves favorably, or, less often, that deliberately defy it. This threatens validity because the measured outcome reflects participants' theories about the experiment rather than the genuine effect of the treatment, a confound especially acute in unblinded laboratory settings where intentions are easy to read. The classic work distinguishing these cues from real treatment effects underscored that almost any feature of the setup — instructions, the experimenter's manner, the apparatus — can leak the expected response. Remedies that matter include blinding participants to condition and hypothesis, using cover stories, deception where ethical, indirect or unobtrusive measures, and post-experimental probes to detect awareness. It is closely related to the Hawthorne effect and social desirability bias, and it matters wherever self-aware participants might perform the experiment rather than simply respond to it.
Where the cues come from
The cues are rarely a single tell; they accumulate from the consent form's framing, the order of tasks, the experimenter's tone, the apparatus on the table, and gossip from earlier participants. Orne's sharpest demonstration was how little it takes to command compliance: asked to add adjacent numbers on sheet after sheet and then tear each completed page into pieces, people persisted for hours at the pointless task simply because it was framed as an experiment. Stepping into the subject role primes people to search for the correct way to behave and to supply it. A within-subjects design, where the same person meets every condition, is especially leaky, because the contrast between conditions all but announces what the study expects to find.
What the evidence shows
Pooled across studies, deliberately planted cues move behavior in the expected direction, but modestly. A 2025 meta-analysis of explicit demand manipulations put the average at Hedges' g = 0.21, a small effect, and found it strikingly heterogeneous: the prediction interval spans from a sizable push toward the hypothesis to a moderate push against it, so demand can inflate or deflate a result. In economics, de Quidt and colleagues deliberately induced strong demand to bound its size across eleven classic tasks and estimated an influence averaging roughly 0.13 standard deviations. The good-subject tendency is real but conditional, strengthening when participants like the experimenter and weakening otherwise. The upshot is a genuine threat, seldom an overwhelming one.
A convenient scapegoat
Because the phrase is easy to invoke, it is often used to wave away inconvenient findings without any evidence that a cue was present or that it biased responses. Sharpe and Whelton, reviewing decades of use, argue the construct has become a rhetorical scarecrow, cited far more than it is measured. Zizzo makes the tighter point: demand threatens validity only when it is positively correlated with the study's true prediction, pushing participants the same way the hypothesis does. Cues that cut across or against the prediction add noise, not bias. Dismissing a result on demand grounds therefore carries a burden of proof, namely to show the cues existed, to show their likely direction, and to show it aligned with the effect being claimed.
Related but distinct
It is worth separating from its neighbors. The Hawthorne effect is reactivity to being observed at all; demand characteristics require the extra step of inferring the hypothesis and acting on it. Social desirability is the wish to look good, which may run with or against the expected answer. Experimenter expectancy effects, documented by Rosenthal, originate in the researcher's own behavior leaking their hopes, whereas demand characteristics live in the participant's reading of the whole situation, cues the experimenter may not intend. A placebo response is a change produced by the expectation of benefit, not by guessing a research aim. The distinctions matter because each has a different remedy: blinding the experimenter fixes expectancy, but only masking the hypothesis touches demand.
Examples
Sensing an experiment is about politeness, participants behave more politely than they otherwise would.
Users in a usability session praise a confusing screen because the designer is sitting beside them. The same screen, tested unattended at home, gets abandoned within seconds.
Volunteers told a drink is being tested for alertness report feeling sharper afterwards. They are reporting their theory of the study back to the researcher, not the effect of the drink.
An employee engagement survey rolled out right after a wellness push draws warm ratings; staff sense which answers management wants and supply them, so the score tracks the campaign's intent, not morale.
In a dictator game, subjects told the study measures generosity give away more, reading the label as an instruction rather than revealing how they would split the money unprompted.
First described in Martin Orne (1962).
Key references
- Coles, N. A., Wyatt, M., & Frank, M. C. (2025). A Meta-Analysis of the Impact and Heterogeneity of Explicit Demand Characteristics. Collabra: Psychology, 11(1), 143005. doi.org/10.1525/collabra.143005
- de Quidt, J., Haushofer, J., & Roth, C. (2018). Measuring and Bounding Experimenter Demand. American Economic Review, 108(11), 3266-3302. doi.org/10.1257/aer.20171330
- Sharpe, D., & Whelton, W. J. (2016). Frightened by an Old Scarecrow: The Remarkable Resilience of Demand Characteristics. Review of General Psychology, 20(4), 349-368. doi.org/10.1037/gpr0000087
- Zizzo, D. J. (2010). Experimenter demand effects in economic experiments. Experimental Economics, 13(1), 75-98. doi.org/10.1007/s10683-009-9230-z
- McCambridge, J., de Bruin, M., & Witton, J. (2012). The Effects of Demand Characteristics on Research Participant Behaviours in Non-Laboratory Settings: A Systematic Review. PLoS ONE, 7(6), e39116. doi.org/10.1371/journal.pone.0039116
- Nichols, A. L., & Maner, J. K. (2008). The Good-Subject Effect: Investigating Participant Demand Characteristics. The Journal of General Psychology, 135(2), 151-166. doi.org/10.3200/GENP.135.2.151-166
- 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. doi.org/10.1037/h0043424