Explainer · Plain-language

What is research bias?

Research bias is any systematic error in how a study is designed, conducted, or reported that pushes findings consistently in one direction, distorting the truth and threatening validity.

CASRAI plain-language explainers — clear answers to recurring research-administration questions

The step most authors miss

Doing CRediT right? Don’t stop at the statement.

A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.

Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.

Systematic error, not random noise

The defining feature of bias is that it is systematic: it nudges results in a consistent direction rather than scattering them randomly. Random error makes individual measurements noisier but tends to cancel out across a large sample; bias does not — it shifts the whole picture and a bigger sample will not fix it. Because biased findings can look clean and precise while being wrong, bias is a serious threat to a study’s validity. Recognising that an error is systematic, and tracing where in the research process it entered, is the first step to controlling it.

Common types of bias

Bias takes many recognised forms. Selection bias arises when the sample is not representative of the population, often because of how participants were recruited or who dropped out. Confirmation bias is the tendency to seek, interpret, or favour evidence that supports a prior expectation. Recall bias occurs when participants remember past events inaccurately or unevenly, common in retrospective studies. Observer (or experimenter) bias is when a researcher’s expectations influence how they record or interpret data. Publication bias describes the tendency for positive or striking results to be published while null findings stay in the file drawer, distorting the wider evidence base.

Guarding against bias

Bias is managed by design and by transparency rather than eliminated entirely. Randomisation and representative sampling counter selection bias; blinding participants and assessors counters observer and expectation effects; standardised, validated measures reduce recall and measurement bias. At the level of the literature, pre-registration and registered reports — where the plan or even acceptance precedes results — directly target confirmation and publication bias, and reporting standards such as PRISMA and the EQUATOR network’s guidelines make methods and outcomes auditable. The aim is not a mythical "bias-free" study but one whose potential biases are anticipated, minimised, and openly disclosed.

Key facts

At a glance

Definition: systematic error that skews findings consistently in one direction
Not: random error, which scatters results and tends to cancel out
Common types: selection, confirmation, recall, observer, publication bias
Effect: threatens validity; a larger sample does not remove it
Defences: randomisation, blinding, validated measures, pre-registration
Standards: PRISMA and EQUATOR guidelines make methods auditable

Common misconceptions

What people often get wrong

Often heard: A large enough sample size cancels out research bias.

Actually: Larger samples reduce random error but not systematic bias. A biased design produces a consistently skewed result no matter how many participants are added.

Often heard: Bias only matters when researchers are dishonest or deliberately cheating.

Actually: Most bias is unintentional, arising from sampling, memory, expectation, or publication patterns. It threatens validity even when everyone acts in good faith, which is why structural safeguards exist.

Often heard: Publication bias is the journal’s problem, not the reader’s concern.

Actually: Publication bias distorts the whole evidence base a reader sees, making effects look stronger than they are. It is a key reason to value pre-registration and systematic reviews that search for unpublished results.

Going deeper