This paper studies abstraction methods suitable to verify very large models of discrete-event systems to be nonconflicting. It compares the observer property to methods known from process algebra, namely to conflict equivalence and observation equivalence. The observer property is shown to be the property that corresponds to conflict equivalence in the case where natural projection is used for abstraction. In this case, the observer property turns out to be the least restrictive condition that can be imposed on natural projection to enable compositional reasoning about conflicts. The observer property is also shown to be closely related to observation equivalence. Several examples and propositions are presented to relate different aspects of these methods of abstraction.