In this course, we will explore a novel approach to integrating structure building with structure modification that may shed more light on the differences between complements, specifiers and adjuncts, especially when it comes to (the direction of) syntactic locality. This will serve not only as an exploration of a concrete set of proposals, but also as a survey of several issues of current interest in syntactic theory, including the nature of the operations Merge and Agree (or valuation), labeling, selectional requirements, the directionality of syntactic dependences, islands and locality.
Standardly, it has been assumed that syntactic structures are created by the operation Merge. Yet, two questions are still subject to further research. The first question concerns labeling. If two elements merge with each other, which of them, if any, becomes the label and why? The second question concerns selection. Why is it that certain syntactic elements may or must merge with an element that carries a particular label? Why is it that determiners must merge with NPs or verbs with their arguments?
In the first part of this course, we discuss these questions in a fair amount of detail and we will explore a particular proposal where labeling and selection go hand in hand. Specifically, we propose that all of the features on two merging elements percolate, except for the selecting and selected features that triggered Merge in the first place. One of the particular outcomes of this proposal is that the difference between complements/specifiers vs. adjuncts falls out naturally: the former are selected by the head they merge with; the latter select their hosts.
The account we set up for structure building allows insights into selective opacity, the fact that certain syntactic domains seem to be transparent for some dependencies, but opaque for others. For example, adjunct clauses are (usually) islands for various types of extraction and phi-agreement, but are transparent for obligatory control.
An important part of our approach is that we derive a clear distinction between structure building, i.e. what is accomplished by Merge, and structure modification. The latter includes things like phi-agreement and (potentially) triggers for movement. We make the claim that these are accomplished by an operation of valuation, which is distinct from, but dependent on the operation of feature percolation / selection that underlies Merge.
Specifically, we hypothesize that the direction of selection and the direction of agreement and extraction go hand in hand in that selection creates pathways for locality. Since adjuncts select their hosts are are not selected by them, they appear opaque to extraction and agreement (where a goal inside an adjunct would value a probe oustide it). However, since adjuncts select their hosts, a probe inside an adjunct may be valued by a goal outside it. This underlies, for instance, obligatory control into adjuncts, where PRO contains the probe and its controller the goal.