Scientific communication takes place at two registers: first, interactions with colleagues in close proximity-members of a network, school of thought or circle; second, depersonalised transactions among a potentially unlimited number of scholars can be involved (e.g., author and readers). The interference between the two registers in the process of peer review produces a drift toward conflict of interest. Three particular cases of peer review are differentiated: journal submissions, grant applications and applications for tenure. The current conflict of interest policies do not cover all these areas. Furthermore, they have a number of flaws, which involves an excessive reliance on scholars' personal integrity. Conflicts of interest could be managed more efficiently if several elements and rules of the judicial process were accepted in science. The analysis relies on both primary and secondary data with a particular focus on Canada.