This paper deals with propositional belief merging. The key problem in this setting is to define the beliefs/goals of a group of agents from a profile of bases, gathering the beliefs/goals of each member of the group. To this aim, a well-studied family of merging operators consists of distance-based ones: the models of the merged base are the closest interpretations to the given profile. Many operators from this family are based on the Hamming distance between interpretations, which can be viewed as a degree of conflict between them. In this paper, we introduce a more general family of merging operators, based on a more primitive concept, namely the conflict between interpretations itself. We show that this family of conflict-based merging operators includes many operators from the literature, both model-based ones and syntax-based ones. We present a number of comparison relations on conflict vectors characterizing operators from this family, and study the logical properties of conflict-based merging operators.