In Progress. Related Slides from Mini-Course at the University of Cambridge: [PDF]
Adjuncts have long been considered syntactic islands (Ross, 1967; Cattell, 1976, a.o.). Prior approaches to adjunct/argument asymmetries have involved according adjuncts some special primitive status: e.g. claiming that they can sideward A-move (Drummond and Hornstein, 2014, etc), or are late-merged (Lebeaux, 1991; Fox, 2002; Abe, 2018), or are phasal (Chomsky, 2008; Müller, 2010). In this paper, we extend a particular approach to structure building and labelling to implement a version of Agree which derives relevant directionality and locality restrictions. We show that this not only derives adjunct opacity but also predicts cases where such opacity is systematically obviated, as in cases of adjunct obligatory control. If the islandhood of adjuncts can be derived from the theory of selection and checking, phase theory is partly unburdened. Note that phases are still required to force successive-cyclic movement and blocking hyperraising to subject/object (Deal, 2017; Halpert, 2019). But by identifying locality phenomena that have a different source, we make it possible to develop a more parsimonious phase theory and potentially even render it superfluous down the line.