Types of Bias (Selection, Recall, Observer, Hawthorne)

MCAT trap: Conflates observer bias (researcher-side) with the Hawthorne effect (participant-side). Observer bias is the researcher's tendency to record findings in line with expectations; the Hawthorne effect is participants changing their behavior because they know they are being observed.

Bias in research methods is heavily tested on the MCAT — and the most important distinction to get right from the start is observer bias versus the Hawthorne effect. Observer bias lives on the researcher's side: the investigator unconsciously records data in a way that confirms their hypothesis. The Hawthorne effect lives on the participant's side: subjects modify their own behavior because they know they're being observed. These require different fixes — blinding the assessor addresses observer bias; having an equally-observed control group partially controls the Hawthorne effect. You'll encounter bias questions in three flavors: pure recall, passage identification, and design critique where you must identify which bias most threatens the conclusion.

What makes this topic hard isn't memorizing the definitions — it's keeping the distinctions clean under pressure. The most commonly confused pair is observer bias versus the Hawthorne effect. Students mash them together as 'observation causes problems,' but they happen on opposite sides of the researcher-participant relationship. Similarly, students reach for blinding as the universal fix for every bias, when in reality blinding targets observer and performance bias, not selection bias. Randomization and allocation concealment handle selection bias. Mixing up the fix is just as wrong as mixing up the bias.

On the MCAT, passage-based questions will describe a study design and bury the bias in a procedural detail — participants self-reporting past exposures, researchers knowing who got the intervention, dropout rates that differ between arms. Train yourself to read every study design detail as a potential bias signal. Ask: who knows what, when is the data collected, and who's missing from the final analysis?

Common misconceptions

Common mistake

Wrong: Observer bias and Hawthorne effect both refer to the researcher's tendency to see what they expect.

Right: Observer bias is the researcher's tendency to record findings in line with expectations; the Hawthorne effect is participants changing their behavior because they know they are being observed.

Observer bias lives on the researcher's side: the investigator unconsciously records, codes, or interprets data in a way that confirms their hypothesis. The Hawthorne effect lives on the participant's side: subjects modify their own behavior simply because they know they're being watched, regardless of what the researcher does. Think of it as researcher-sees-what-they-expect versus participant-acts-differently-when-watched. These require different fixes — blinding the assessor addresses observer bias; the Hawthorne effect is partially controlled by having a control group that is equally observed.

Common mistake

Wrong: Blinding prevents selection bias.

Right: Blinding prevents observer and performance bias; selection bias is prevented by randomization and proper allocation concealment.

Blinding happens after participants are already enrolled — it controls what researchers and participants know about group assignment during the study. Selection bias happens at enrollment, when the groups differ in a systematic way before any intervention occurs. No amount of blinding can fix a sample that was already skewed. Randomization with proper allocation concealment (hiding the assignment sequence until the moment of enrollment) is the tool that prevents selection bias.

Common mistake

Wrong: Prospective cohort studies are susceptible to recall bias because participants must remember their exposures.

Right: Prospective cohort studies record exposures before outcomes occur, eliminating recall bias; recall bias is a concern in retrospective designs like case-control studies.

In a prospective cohort study, exposures are measured and recorded at the start, before any outcome has developed — so there's nothing to 'remember' inaccurately. Recall bias is a retrospective problem: in case-control studies, participants with disease are more likely to dredge up past exposures than healthy controls, creating asymmetric misclassification. If a question mentions participants are asked to remember past behavior, that's your signal for a retrospective design and recall bias risk.

Common mistake

Gap: Unaware that differential dropout (attrition bias) can systematically distort RCT results and that intention-to-treat analysis addresses it

Attrition bias occurs when participants who drop out of a study differ systematically from those who remain, and intention-to-treat analysis is the primary method to mitigate it in RCTs.

Attrition bias occurs when dropout isn't random — the people who leave one arm of a trial differ systematically from those who leave the other arm (or stay). For example, if participants in a treatment group who are experiencing side effects disproportionately drop out, the remaining treatment group looks artificially healthier than it really is, inflating the apparent benefit. Intention-to-treat (ITT) analysis combats this by analyzing all participants in their originally assigned groups regardless of whether they completed the study, preserving the protection that randomization provided at the start.

Free Deck audit

See if your Anki deck covers this topic.

Upload your deck →

Guided session

Stuck on this? An AI tutor that probes your understanding.

Start a session →

What the exam tests

Recognize each major bias type — selection, recall, observer, Hawthorne effect, social desirability, attrition, and publication bias — from a brief description or study scenario.
Match each bias to its correct mitigation strategy: randomization and allocation concealment for selection bias, blinding for observer/performance bias, prospective design for recall bias, and intention-to-treat analysis for attrition bias.
Read a passage describing a study's methodology and identify the single dominant bias operating in that design, then propose the specific design change that would address it.
Evaluate a flawed study that is vulnerable to multiple biases simultaneously, determine which bias most directly threatens the study's main conclusion, and explain the direction of distortion it would cause.

Can you avoid these mistakes?

A case-control study asks lung cancer patients and healthy controls to report their lifetime smoking history. What bias is most likely, why does it apply here but not in a cohort study, and what would you see in the data if it were operating?

An RCT of a new antidepressant shows impressive results, but 30% of the treatment arm dropped out due to nausea while only 5% of the placebo arm dropped out. The researchers analyze only participants who completed the trial. Name the bias, explain which direction it distorts the result, and state the analysis method that should have been used.

A researcher studying hand-washing compliance observes nurses directly during shifts and records whether they wash hands between patients. Compliance is 95%. A separate study using hidden sensors in the same hospital records 60% compliance. Which bias explains the discrepancy, and is it the researcher or the participant driving the error?

A pharma company runs four trials of a drug; three show no effect and one shows a significant benefit. Only the positive trial gets published. What bias does this create, in which direction does it skew the apparent evidence base, and what systematic review technique is designed to detect it?

Types of Bias (Selection, Recall, Observer, Hawthorne)

Common misconceptions

What the exam tests

Can you avoid these mistakes?

Related topics