Concatenation and Grammar: How to Get Systematic
Dr. Paul M Pietroski
Tuesday, March 28, 2006, 01:00pm - 02:00pm
University of Maryland, Department of Linguistics and Philosophy
Human languages exhibit recursivity, in the form of adjunction, and grammatical relations that hold between predicates and their arguments. Simple cases of adjunction�as in �red ball� and �sang loudly��can be analyzed, to a first approximation, as predicate conjunction. Letting �e� range over entities and events: �Red(e) & Ball(e)�, �PastSinging(e) & Loud(e)�, etc. But a noun or verb can be modified by many adjuncts, while a verb takes only a small number of grammatical arguments; and the phrase �stabbed Caesar� cannot be analyzed as �PastStab(e) & Caesar(e)�. It can, however, be analyzed as �PastStab(e) & Patient(e, Caesar)�; where the Patient of a stab is the entity stabbed, and the grammatical argument corresponds to a participation relation�being the Patient of�that an entity can bear to an event. These increasingly standard remarks invite a two-part speculation: (i) the recursive core of natural language is a system for concatenating predicates, with concatenation signifying predicate-conjunction; and (ii) grammatical relations, like being the subject or object of a verb, reflect an additional �thematic� component that makes it possible to combine predicates with designators in a suitably general way. The idea is that while the name �Caesar� is not a predicate that can be coherently conjoined with the verb �stabbed�, being an argument of a verb matters semantically; and the complex predicate, �Patient(e, Caesar)� is conjoinable with the event predicate �PastStab(e)�. From this perspective: grammatical relations let humans use the recursive core of language to create complex conjunctive representations from constituents that are unconjoinable; and lexicalization is a process in which prelinguistic representations are associated with suitably general predicates, like �PastStab(e)�, that are conjoinable with others. I think this conception of semantics has many virtues, not least of which is that it fits nicely with the idea that linguistic meanings are �Concept Construction Instructions� that let us humans systematize our prior representations�in roughly the ways that Spelke and others have recently been suggesting.
Note: If you would like to receive email announcements about the colloquium series, please contact the Business Office to have your name added to our announce lists at firstname.lastname@example.org.