Selection bias: ask who entered the evidence and who did not

For a reader, that prompts a question before interpreting the result: how did these people or observations become part of the evidence? We use an invented community example, not a judgment about any real study or a claim that a particular survey is invalid.

The people at the meeting are not everyone

Imagine a club asking people at its Tuesday afternoon meeting whether Tuesday afternoon is a convenient time. Most attendees say yes. The organizer writes, “Tuesday afternoons suit all our members.”

The replies tell us what the people answering said. They do not include members who could not attend that afternoon. The same feature being discussed—the meeting time—also shaped who was present to give an answer. A reader can notice that gap without deciding what the absent members would have said.

An honest account of the invented result would identify the respondents as attendees at that meeting. The club could then decide how to seek other views, but this page does not prescribe a scientifically representative survey or claim that a second message would solve every selection problem.

Trace the route into the result

Who the conclusion appears to describe:
Who was eligible or invited to contribute:
How they were approached or selected:
Who actually contributed:
Groups or circumstances the process may have missed:
What the authors report about this limitation:
What remains unknown:

Do not fill “unknown” with a story about people's motives. Not replying does not tell us whether someone agreed, disagreed, overlooked the request, or could not access it.

Distinguish different questions

Sample size asks how much information was collected. Selection asks how it entered the record. Counting more replies from the same Tuesday meeting would not give the invented club direct answers from people absent that day.

Selection bias is also not a label to attach to any result you dislike. Read the actual recruitment and selection description and the authors' limitations. If you cannot access that material, report the access limit rather than declaring the methods checked.

Make room for more than one viewpoint

A child can notice that their sibling was not asked. A parent might ask whether a headline concerns people with similar ages or circumstances. A partner can help locate the recruitment section without claiming to resolve the study's validity. A grandparent's account may add a perspective while remaining one person's account, not a replacement sample.

For ADHD research, this glossary entry is a reading aid. It does not establish whether a clinical finding applies to a particular person.

Source

[1] National Library of Medicine. Selection Bias: MeSH descriptor D015983. MeSH 2026. Record last updated July 12, 2017; accessed September 5, 2026. The scope note supports [1]. This is terminology guidance, not a validation of Ready's invented meeting example or a full methods review.

Related guides

• Sample size

  : separate the number of observations from their selection.

• Observational study

  : identify one kind of study design.

• Confounding

  : distinguish selection from other unresolved differences.

• Primary source

  : find the original methods description.