Publication bias, the phenomenon in which studies with positive results are more likely to be published than studies with negative results, is a serious problem in the interpretation of scientific research. Various hypothetical models have been studied which clarify the potential for bias and highlight characteristics which make a study especially susceptible to bias. Empirical investigations have supported the hypothesis that bias exists and have provided a quantitative assessment of the magnitude of the problem. The use of meta-analysis as a research tool has focused attention on the issue, since naive methodologies in this area are especially susceptible to bias. In this paper we review the available research, discuss alternative suggestions for conducting unbiased meta-analysis and suggest some scientific policy measures which could improve the quality of published data in the long term.