Publication bias occurs when one type of study result is more likely to be published than another.
For example, publishing results of studies that show a new treatment provides significant benefits, but not always publishing studies that show less or no benefit, will lead to publication bias. The effect is to exaggerate the benefit of the new treatment.
Methods exist to help address publication bias. For example, there are public databases where clinical trials can be registered before they begin. Thus, all the registered trials are known, including those that eventually demonstrate no significant benefit from the study treatment. Researchers are encouraged to register their trials because, for example, certain high-profile medical journals insist that they will only publish trials previously registered on a public database.