week of september 3

Lecture 1: combining meanings. We talked about what I called "loose" semantic composition, as exemplified by N-N compounds, in contrast to "tight" composition of the kind that comes with phrasal syntax. How could we characterize the difference?

Lecture 2: First steps towards Fregean meaning composition. We thought about the semantic contribution of verbs. We began by trying to isolate the contribution of verb roots, that is, we tried to abstract away from the possible contributions of aspect, tense, mood, and whatever pieces of verbal inflection may make their own compositional contribution to the meaning of a verb. We then looked at the traditional proposal to construe the denotation of (the root of) an intransitive verb like "procrastinate" as the set of all procrastinators. We acknowledged a first gut reaction of dissatisfaction with this proposal. Why should "procrastinate" pick out the set of all procrastinators, rather than the set of all instances of procrastination, for example? A serious evaluation was promised for the following week. In the meantime, we explored the consequences of the traditional proposal for an account of meaning composition. We first had to get clear about what the meaning of a sentence could be: to know the meaning of a sentence is to know its truth-conditions. For example, a person who knows the meaning of "Emma procrastinates" knows that the sentence is true just in case Emma procrastinates. A semantic theory should be able to derive these truth-conditions compositionally from suitable denotations for the parts of the sentence. "Emma" is a proper name, and should thus denote the individual Emma. The verb "procrastinates" denotes the set of all procrastinators according to the proposal we are exploring. We can then say that "Emma procrastinates" is true just in case Emma is a member of the set of procrastinators. More generally, for any referential DP, a sentence of the for DP VP is true just in case the denotation of the DP is a member of the denotation of the VP. This last generalization about how to compute the truth-conditions of a simple sentence from the denotations of its parts contains a requirement about the relation between the DP and the VP denotation: it has to be set membership in this case. What exactly is the status of this requirement? Where does it come from? Is it attached to particular syntactic structures? At that point, I introduced a more general vision about semantic composition that goes back to the logician Gottlob Frege. Here is a simple-minded way of describing Frege's vision: The meaningful expressions of natural languages denote either individuals or functions, including functions that can take other functions as arguments. All meaning composition might then reduce to Functional Application. The handout for this lecture illustrates the proposal for our example sentence.

To think about: Is Frege's vision, as I described it, just a technical trick or an empirically contentful proposal? Might further assumptions be necessary or desirable to turn it into an interesting proposal? What could those assumptions be? Check out Manfred Krifka's short entry on compositionality in the MIT Encyclopedia of the Cognitive Sciences. (The links require access from the UMass campus. If you are not on campus, you can access the MIT Encyclopedia by logging into the CogNet data base via the UMass library website). A longer philosophical discussion of compositionality can be found in the Stanford Encyclopedia of Philosophy (by Zoltán Gendler Szabó). A very accessible article on compositionality that is at the same time an introduction to semantic analysis from a linguist's point of view is Barbara Partee's "Lexical Semantics and Compositionality", which appears in the 1995 edition of An Invitation to Cognitive Science (D. Osherson general editor); Volume 1: Language, edited by L. Gleitman & M. Liberman, Cambridge/Mass. (the MIT Press), 311 - 360.