Publication bias
Publication bias happens when the studies that become visible in the literature differ systematically from all the studies that were actually conducted.
What publication bias is
Publication bias is selective visibility in the research record. It occurs when studies are more or less likely to be published because of their results, direction, size, novelty, or perceived interest. If published studies are not representative of all studies that were done, readers may get a distorted picture of the evidence.
Why null results disappear
A study that finds no clear effect can be harder to write up, harder to place in a journal, and less rewarding for careers. Researchers may decide not to submit it, journals may prefer more surprising findings, and funders or sponsors may have weak incentives to highlight disappointing results. Over time, missing null studies can make effects look stronger than they are.
How it affects evidence
Publication bias is especially important in systematic reviews and meta-analyses. These methods combine published studies to estimate an overall effect. If the available studies overrepresent positive findings, the combined estimate can exaggerate benefits, understate harms, or make uncertain evidence look settled.
Funnel plots
A funnel plot is a scatter plot used in meta-analysis to compare study effects with study size or precision. A visibly asymmetric plot can suggest that smaller studies with certain results are missing. Funnel plots are useful clues, not proof: asymmetry can also come from differences in methods, populations, quality, or real study effects.
Ways to reduce it
Researchers and institutions can reduce publication bias by registering studies before results are known, reporting all planned outcomes, sharing protocols and data, publishing null results, using registered reports, and requiring clinical-trial results to be posted in public registries. Reviewers can search gray literature and registries, not only journal articles.
Limits of detection
Publication bias is difficult to measure because the missing studies are, by definition, hard to see. Statistical tests and visual tools can warn reviewers, but they cannot fully reconstruct unpublished work. Good prevention is often stronger than after-the-fact correction.
Why it matters
Evidence should reflect what researchers found, not only what looked publishable. Publication bias matters because selective visibility can waste resources, mislead decisions, and make weak claims appear stronger than they really are.