The empirical facts about right-node raising (RNR) lead to fundamentally conflicting analytical conclusions. There is strong evidence that RNR does not obey syntactic constraints of any kind, which in turn suggests that RNR is not a syntactic operation, but there is also evidence that strongly favors a syntactic analysis. The idiosyncratic and almost paradoxical nature of the phenomena indicates that no simple unified analysis of RNR can be formulated. In order to resolve this empirical and theoretical impasse, I propose that what is usually called RNR is best seen as the conflation of three completely unrelated kinds of phenomena: VP/N′-ellipsis , extraposition , and (backward ) periphery deletion . Although they are fundamentally different, these phenomena can yield structures that are superficially similar and, in some cases, apply to the same strings. The latter is one of the major factors that has misled previous accounts. Once this three-way confound between ellipsis, extraposition, and deletion is taken into account, the contradictory idiosyncrasies about RNR vanish, and a wide range of cases are obtained as predictions of independently motivated accounts of VP/N′-ellipsis and ATB extraposition phenomena. This article offers an explicit formalization of the phenomena under discussion in sign-based construction grammar (), a framework that combines insights from head-driven phrase structure grammar (HPSG; ) and Berkeley construction grammar ().