A thorough assessment of this is available on LITFL. In short, it is the bias which creeps in not during study design or publication but afterwards- when the reader finishes the paper, and interprets its findings.
Difference in the interpretation of evidence which supports your opinion, as compared to evidence whih challenges it.
Discounting a study by finding selective faults within it
A decrease in scepticism produced by credible-sounding scientific foundations for the study
The hypothesis itself introduces prejudice and determines experimental outcome