The tendency is strong to take the notion of "conflict of interests" for granted as if it had an invariant meaning and an ethical content independent of the historical context. It is doubtful however, from an historical and sociological point of view, that many of the cases now considered as instances of "conflicts of interests" would also have been conceived and perceived as such in, say, the 1930s. The idea of a "conflict of interests" presupposes that there are indeed interests in conflict. Conversely, as long as there is a consensus among the different groups involved, they will not conceive and even less denounce a given practice as being an instance of a "conflict of interests". In this article we will show that the content of the discussions over conflicts of interests has changed over time in close relation with the transformations of the research system. In other words: there are social conditions for the emergence of "conflicts of interests". The changing meaning of the notion is assessed by analyzing the presence of the expression "conflicts of interests" in the magazine Science over the past century. Three different meanings emerge and their content has evolved in close link with the changing structure of the relations between the scientific community first with the State and then with industry. It moved from a situation external to the scientific community to a debate going on inside the scientific community generated by the growing relations between university and industries.