Questions: Experimenter Bias and Expectancy Effects

Question 1

A researcher tests whether a new therapy reduces anxiety. She administers the therapy in person to participants and knows which group is receiving treatment versus placebo. Her assessment of outcomes relies partly on clinical observation. Why is this design problematic?

• A. The study uses a placebo, which is unethical in psychological research
• B. The researcher's knowledge of group assignment may cause her to unconsciously behave differently toward participants or interpret ambiguous outcomes in ways that favor the treatment group
• C. Clinical observation is never a valid outcome measure in psychological research
• D. Random assignment is impossible in therapy research, which invalidates the findings

Answer: (B)

This design is vulnerable to experimenter bias: the researcher's expectations can influence how she delivers therapy (e.g., more warmth, more encouragement for the treatment group), how she interprets ambiguous clinical observations, and how she records outcomes. Crucially, these effects operate through subtle, often unconscious channels — the researcher does not need to intend to bias the results. The Rosenthal rat studies showed that even handling differences too subtle to observe directly could produce different outcomes. The solution is blinding the researcher to group assignment.

Question 2

Which combination of controls most directly addresses both experimenter bias and demand characteristics simultaneously?

• A. Large sample size and random assignment to conditions
• B. Double-blind design combined with standardized, scripted procedures
• C. Pre-registration and replication by independent labs
• D. IRB oversight and thorough informed consent procedures

Answer: (B)

A double-blind design prevents experimenters from knowing group assignments (targeting experimenter bias) and prevents participants from knowing their condition (targeting demand characteristics). Standardized procedures further reduce variability in how the experimenter interacts with participants, eliminating a key transmission channel for unconscious expectancy effects. The other options are valuable for other reasons: random assignment addresses confounding, pre-registration prevents HARKing, replication establishes generalizability — but none simultaneously targets both experimenter and participant-side bias.

Question 3

Experimenter bias mainly occurs when researchers deliberately manipulate or fabricate data to support their hypotheses.

• T. True
• F. False

Answer: False

Experimenter bias is defined by its unconscious, unintentional nature — which is what makes it so insidious and so different from fraud. In Rosenthal's Pygmalion study, teachers genuinely believed they were treating all students equally, yet they provided more warmth, challenge, and feedback to 'late bloomers.' Similarly, experimenters who 'knew' their rats were bright unconsciously handled them differently. The bias operates through micro-behaviors that neither the experimenter nor the participant may notice. This is precisely why procedural safeguards like blinding are necessary — self-monitoring alone cannot prevent what you're not aware of.

Question 4

Demand characteristics can cause participants to produce results that either confirm OR disconfirm the study's hypothesis, depending on the participant's motivations.

• T. True
• F. False

Answer: True

Participants who infer a study's hypothesis have two common response strategies: 'playing along' (cooperating with what seems expected) or 'screw you' responding (deliberately producing contrary results to resist being manipulated or to appear unique). Both responses threaten internal validity, just in opposite directions. A cooperative participant inflates effect sizes; an oppositional participant deflates them. This is why demand characteristics are treated as a bias regardless of direction — the problem is that responses are shaped by perceived expectations rather than the actual independent variable.

Question 5

Explain how the Pygmalion study demonstrates that experimenter bias can operate without any deliberate intent to distort results, and what this implies for experimental design.

Model answer: In the Pygmalion study, teachers were randomly told that certain students were 'late bloomers.' By year-end, those students showed greater IQ gains — not because teachers consciously gave them advantages, but because subtle behavioral differences (more warmth, more challenging material, more positive feedback) accumulated over time without teachers realizing it. This shows that expectancy effects operate through unconscious behavioral channels. For experimental design, the implication is that good intentions are insufficient: even honest, motivated researchers must be blinded to group assignments to prevent their expectations from influencing outcomes.

The mechanism — unconscious transmission through subtle behavioral cues — is what makes experimenter bias methodologically serious. If it only occurred through deliberate fraud, peer review and replication would catch it. Because it operates below the level of awareness, it can replicate across labs with equally honest researchers all sharing the same theoretical expectations. This is one reason that blinding is not just a best practice but a methodological necessity, and why independent replication (by researchers with no stake in the original finding) is the final safeguard.