Division of Labor in Right Node Raising Constructions

Abstract

The phenomenon of Right Node Raising (RNR) has invited a wealth of discussion in the literature, centered around the following three major approaches: rightward across-the-board extraction, backward ellipsis, and multi-dominance. Despite the success of each of these approaches to some substantial degree, however, it still remains agnostic whether all the reported properties of RNR can be captured by any single approach exclusively. This paper argues that in deriving RNR structures, both parallel merge and (simple) merge should be operative for the optimal derivation of RNR, assuming that UG employs three types of merge, i.e., external merge, internal merge, and parallel merge (sharing by multi-dominance) (Citko 2005). I will provide empirical evidence for the division of labor between the two operations in RNR, and propose that given the interface economy, parallel merge is the most preferred option for RNR constructions; but at the same time, merge (followed by ellipsis), another independent operation in UG, should be a readily available option for RNR constructions.