When one only publishes particular (usually favourable!) results. This may be as a deliberate policy, but may also be due to the policy of journals only accepting positive results. The impact of selective publishing is that many results, in particular negative results, are never published, a phenomenon known as the file drawer effect. Because of this the number of false positives in the published literature may be far higher than one would expect given the quoted significance level of results, one of the causes of the statistical crisis.
Defined on page 82
Used on page 82