Locality and the argument-adjunct distinction:Structure-building vs. structure-enrichment

点击翻译按钮获取中文摘要

Syntax is partly about structure-building: the combination of words into phrases and phrases into sentences. But it's also about dependencies between different elements in a sentence. We aim to compare how "local" different types of syntactic dependencies are in a range of languages (that is, how much syntactic structure they can traverse), and build a theory to explain the different locality patterns that we uncover.For example, in the Hindi sentence "Vivek-ne kitab parh-nii chaah-ii" ("Vivek wanted to read the book"), "parh-nii" ("read") and "chaah-ii" ("want") show feminine inflection, because they agree with the feminine object "kitab" ("book"). This "long-distance agreement" (where the object of a subordinate clause influences the shape of multiple verb forms) is an example of a syntactic dependency.Another nonlocal dependency is "control", where a noun phrase in one clause determines the interpretation of an implicit argument in another clause. In the same Hindi example above, "Vivek" is interpreted not only as the subject of "want" but also as the subject of "read". Again, this dependency is syntactically mediated.Different syntactic dependencies permit different amounts of nonlocality. We compare different dependency types and ask whether they can cross adjunct boundaries, where an adjunct is an unselected optional phrase within the sentence (like "in the garden" in "I was walking in the garden"). Control dependencies clearly can cross adjunct boundaries: in "They left after eating", the adjunct is the optional modifier "after eating", and the implicit subject of "eating" is controlled by "they". However, long-distance agreement appears to be impossible across an adjunct boundary: to our knowledge, no language allows a pattern like "Vivek left.FEM after buying.FEM the book.FEM", where long-distance agreement between "left" and "book" crosses an adjunct boundary. Other types of syntactic dependency show more complex patterns, sensitive to the specific type of adjunct and the nature of the elements that the dependency relates.Our aim is to describe and explain these patterns. Currently, no theory explains why different types of dependency interact with adjunction in these different ways. We believe that the different patterns can be explained in terms of two new ideas: (1) a particular conception of the way that adjuncts relate to their hosts, and how it differs from e.g. the relationship between a verb and its object; (2) a theory that accounts for interactions between different types of syntactic dependencies in terms of how the structure-building mechanism determines when agreement and control can take place.