Cooperating constructions

Robert Malouf

University of Groningen


Mismatch phenomena have long been of central importance in the development of constraint-based theories of grammar (such as Lexical-Functional Grammar, Head-Driven Phrase Structure Grammar, Construction Grammar, and others). While the details differ greatly, each of these theories posits independent representations for (at least) syntactic, semantic, and morphological structure which exist in parallel and are mutually constrained by the grammar.

Recently, the introduction of Optimality Theoretic constraint interactions has shed new light on the place of structural mismatches in constraint-based grammars. OT introduces two key notions, constraint ranking and competition among forms, to account for the fact that under certain circumstances forms may violate grammatical constraints. In particular, a form may violate a constraint so long as every competing form which does not violate it does violate another more highly ranked constraint. So, under OT, mismatches arise when a form can only satisfy a highly ranked constraint by violating a lower ranked constraint which enforces a correspondence between levels.

While many analytic insights have come out of the introduction of OT ideas to constraint-based grammar, the fit between the two traditions is not entirely comfortable. In many respects, OT (at least as it is usually talked about) is incompatible with the fundamental design principles of constraint-based grammar. One of the underlying goals of constraint-based theories, which distinguishes them from approaches to generative grammar descended from the transformational tradition, is to construct a theory of linguistic competence which is compatible with what is known about human language processing. To this end, constraint grammars are strictly declarative sets of constraints, which impose no inherent bias towards either production or comprehension, allow many different kinds of information to be integrated in parallel during processing, and do not require the language user to perform any psychologically implausible operations. OT on the other hand introduces to the grammar a production-oriented procedural metaphor which requires the evaluation of an infinite set of candidate forms.

In this paper, I suggest an alternative view of constraint interaction that allows one to capture many of the key insights of OT while still preserving the methodological goals of constraint-based grammar formalisms. Rather than setting up a competition model which resolves conflicts between constraints on a form by form basis, we can view the inventory of constructions as a set of mutually cooperating constraints. Every form must be an instance of one or more constructions, so the set of possible constructions determines which forms are valid and which are not. Constraint conflicts can be resolved at the level of the construction, allowing OT-like constraint interactions while preserving the strictly monotonic licensing relationship between forms and constraints.