This article explores the phenomenon of pied-piping within the context of two assumptions. The first assumption is that a realistic natural language parser must assign incoming words incrementally into the syntactic representation as it is built during the parsing process. The second assumption is that this process makes use of Merge, the core recursive operation in syntax. These assumptions lead us to consider an extension of the standard theory of Merge, in which phrase structure is derived from a linear string of words in a left-to-right/top-down fashion (e.g. Phillips, 1996, 2003). We propose a generalization of the Phillips architecture that describes and correctly parses simple and complex pied-piping constructions in three languages: English, Italian and Finnish. We then show that the proposed hypothesis derives (without any additional assumptions) several well-known constraints to pied-piping and operator movement, such as Condition on Extraction Domains (CED), the Left Branch Condition, the extended projection principle (EPP), the complex properties of picture nouns and certain differences between relative and interrogative clauses. A computational implementation and full formalization of the hypothesis together with a test corpus is used to demonstrate the viability of the approach. We conclude that the top-down approach provides a useful tool for grammatical analysis.
Read full abstract