Ivan A. Sag: Construction Grammar Research

Since the mid-1980s, research in HPSG has used typed feature structures to model both words and phrases, as two varieties of the `sign'. This leads naturally to the view that phrases (like words) are organized into families, where the `family resemblance' is captured by a constraint on a common supertype. Carl Pollard and I took steps toward developing this conception of grammar in our 1987 and 1994 books on HPSG, but it was only in the early 1990s that I realized how fine-grained the hierarchies should be and how well-suited typed feature structures are for formalizing the basic intuitions of a construction-based approach to grammar. My first attempt at working out a partial theory of this, with respect to English relative clauses, was in the following paper:

Sag, Ivan A. 1997. English Relative Clause Constructions. ( .ps file. .pdf file). Journal of Linguistics 33.2: 431--484.

More recently, Jonathan Ginzburg and I explored this approach in a much more comprehensive way. Our collaborative investigation of the syntax and semantics of English interrogative constructions, which began in 1996, led to the publication of our 2000 book:

Ginzburg, Jonathan and Ivan A. Sag. 2000. Interrogative Investigations: the form, meaning and use of English interrogative constructions

On-line excerpts from this book can also be found in the website for my course Linguistics 221A -- Foundations of English Grammar. This conception of grammar forms the basis for the design of the English Resource Grammar, developed at CSLI's LinGO Lab.

Finally, I've been working on the issue of Locality. Constructional and selectional phenomena are local in nature, though most theories of grammar do not predict this fact. For example, verbs select for the category of elements they combine with directly, not for the category of elements contained within those elements. Similarly, constructions, when properly analyzed, involve constraints that relate mothers and daughters (as in Context-Free Grammar), but none that directly relate mothers and `granddaughters'. I am currently developing a theory of both notions of locality. Here is a handout from a talk I gave about this at the 2001 HPSG Conference in Trondheim: (.ps file).

Comments?
Last modified: Sat Feb 9:12:55 PST 2002