The researcher is careful, experienced, and committed to good science. They design the study thoughtfully, collect the data honestly, and write up the results in good faith. The findings confirm the hypothesis.
What they may not have noticed: expectation shaped the study at every stage - in how ambiguous responses were coded, in which borderline data points were flagged as outliers, in how participants were treated in the two conditions, in which results made it into the final analysis. Nothing was fabricated. The bias operated without intention.
This is experimenter's bias - also called expectation bias - the tendency for researchers to favor, accept, and publish data that match their expectations, while discounting or discarding data that contradict those expectations.
What Experimenter's Bias Is
Experimenter's bias is a systematic distortion that occurs when a researcher's expectations about the outcome of a study influence the study's results. The bias does not require dishonesty or fraud. It operates through dozens of small, often unconscious decisions that accumulate to produce results tilted in the direction the researcher anticipated.
The bias is particularly insidious because it is invisible to the person producing it. The researcher experiences their decisions as legitimate judgment calls - and many of them are. The problem is that those calls are not being made blindly: they are being made by someone who knows what they hope to find.
How the Mechanism Works
Four pathways allow experimenter expectations to enter results:
Interpersonal cues. When researchers interact with participants directly - giving instructions, asking questions, conducting assessments - they can unconsciously signal what responses are expected or desired. Participants pick up on tone, pace, and phrasing in ways that shift their behavior toward the expected outcome.
Selective attention. Data collection involves constant judgment about what counts. When a response is ambiguous, a researcher's expectation determines which category it lands in. These micro-decisions favor the expected pattern across thousands of observations.
Ambiguous interpretation. When results are mixed or uncertain, they tend to be interpreted in whichever direction confirms the hypothesis. The same data point that would be cited as evidence of the effect by a believer is cited as noise by a skeptic.
Post-hoc decisions. Choices about outlier exclusion, analysis strategy, and which conditions to include in the final write-up are often made after the researcher has seen the data. This makes them susceptible to motivated reasoning, even when the researcher is confident they are being objective.
Two Recognisable Scenarios
In research: A clinical researcher testing whether a new training programme improves stress responses believes strongly in the programme. During assessment sessions, they are slightly more encouraging with the treatment group - not deliberately, but perceptibly. Borderline readings favour the expected outcome at coding. Two outliers from the control group are removed at analysis. The final result shows a significant effect. The researcher did not lie. Every individual decision was defensible. The cumulative effect was a result shaped by what they hoped to find.
In everyday judgement: Someone is deciding whether a difficult colleague is genuinely trying to change or just performing effort to avoid consequences. They hold a prior view. In subsequent interactions, the moments that confirm their expectation register clearly; the moments that contradict it get filed as exceptions, edge cases, or attempts at manipulation. The story they tell about that person - and act on - is assembled from selectively weighted evidence. This is experimenter bias outside the lab: expectation shaping what counts as proof.
Why It Matters
Experimenter bias threatens the validity and reproducibility of scientific findings. When bias systematically inflates effect sizes or creates false positives, it can mislead subsequent research, waste resources on follow-up studies, and erode trust in science. In applied fields - medicine, education, policy - research shaped by experimenter expectation can influence real decisions about people's lives.
The bias is also self-reinforcing: once a finding exists in the literature, researchers building on it are anchored to it, and their own expectations are shaped by the published result.
The Common Misunderstanding
Two myths make experimenter's bias harder to recognise and address.
The first is that the bias only affects careless or dishonest researchers. In reality, well-intentioned, rigorous researchers are fully susceptible. Expectation operates below the level of deliberate choice, and good intentions do not disable it.
The second is that the bias is confined to psychology or the soft sciences. Experimenter effects have been documented across disciplines wherever humans make judgment calls during data collection, coding, or analysis - which is nearly everywhere.