Demand characteristics are cues in the research environment that communicate to participants what behavior or response is expected, leading them to modify their behavior to align with perceived experimental hypotheses. Participants may attempt to help the experimenter confirm predictions, demonstrate competence, or respond in socially desirable ways. These effects can artificially inflate or deflate treatment effects and represent threats to internal and construct validity. Techniques such as blind procedures, deception, cover stories, and plausible alternative explanations help minimize demand characteristic effects.
From experimental research design, you know that the logic of a controlled experiment depends on isolating the independent variable as the only systematic difference between conditions. You've also worked through why control and experimental groups must be treated identically except for the treatment itself. Demand characteristics represent a threat that originates inside the participant — from their own psychology — rather than from flaws in the experimental apparatus.
Research participants are not passive measurement instruments. They are curious, socially motivated people who enter a study already generating their own hypotheses about what the researcher is investigating — and they often try to act in ways consistent with those hypotheses. Demand characteristics are the totality of cues in the research context that signal what response is expected: the experimenter's tone, the wording of instructions, the equipment visible in the room, the institutional affiliation, even the phrasing of the consent form. When participants detect these cues and adjust their behavior accordingly, the independent variable is no longer the only thing that differs between conditions — participants' beliefs about the study are doing additional work.
Martin Orne, who gave demand characteristics their name, observed that participants in psychology research are fundamentally in a cooperative relationship with researchers. They want to help produce a successful study. This creates what Orne called the "good subject" effect: participants act in ways they believe will confirm the study's hypothesis, even when doing so requires departing from their genuine reactions. But demand characteristics operate in more than one direction. Some participants deliberately oppose perceived expectations (the "screw you" effect), and others respond in socially desirable ways that have nothing to do with the manipulation at all. All three patterns contaminate the signal the experiment was designed to measure.
This is why blind procedures are standard in rigorous experimental research. In a single-blind design, participants do not know which condition they are in, severing their ability to behave strategically in relation to the treatment. In a double-blind design, neither participants nor experimenters know condition assignment, also preventing experimenters from inadvertently communicating expectations through subtle behavioral cues — which is the related problem of experimenter bias. Cover stories and deception serve the same purpose: if participants believe they are in a study about memory when the real manipulation involves social exclusion, demand characteristics relevant to the actual manipulation cannot operate because participants don't know what behavior is being "demanded."
Post-experimental inquiry — systematically asking participants after data collection what they thought the study was about, and whether they suspected the true hypothesis — is the primary method for detecting demand characteristic contamination. If a substantial proportion guessed the hypothesis and those participants show larger effects, demand characteristics are likely inflating the treatment estimate. The appropriate response is to re-run analyses excluding hypothesis-aware participants and report whether the effect survives, not to omit the inquiry results. This kind of sensitivity analysis is part of transparent reporting and allows readers to judge how much of the effect is genuine versus artifact.