Definition

Publication bias

The tendency for significant, positive results to get published while null results stay in file drawers — skewing what the literature shows.

If positive findings are more publishable, the published record overstates effects: meta-analyses inherit the inflation, and replication attempts disappoint. Funnel-plot asymmetry and related tests diagnose it; the file-drawer problem names it.

Structural fixes are gaining ground: preregistration, registered reports (accepted before results exist), trial registries, and journals that welcome null results. Practical implication for readers: treat published effect sizes as upper estimates, especially from small-study literatures.