On the structural representation of silence

One of the most prominent and widespread features of human language is that much of what we communicate is actually left unspoken. The linguist's job is to understand how this silent content interacts with linguistic utterances. In this talk, I consider several cases of constituent ellipsis like the English examples in (1) and (2):

1a. utterance: I know Sam saw someone, but I don't know who.
1b. interpretation: I know Sam saw someone but I don't know who he saw.

2a. utterance: If you can go to this amazing performance, you should.
2b. interpretation: If you can go to this amazing performance, you should go to this amazing performance.

Two observations about such structures seem fairly straightforward: first, there are structural constraints on what can be omitted; for example, (2c) is not grammatical, hence the * notation:

2c. *If you can go to this amazing performance, you.

Second, listeners know how to interpret the missing material because it is in some sense identical to something that was already uttered in the discourse. In sum, the silence has linguistic content of some kind.

All this leads to the question of how we should characterize the nature of this silent linguistic content. A natural guess is that we call up a semantic representation of the missing material. However, drawing on evidence from a range of languages, I present an argument in favor of the surprising claim that the representation associated with the unpronounced material is at least in part syntactic. This predicts --- correctly, I show --- that the licensing constraints on ellipsis constructions should be at least partly syntactic as well.

Speaker: Vera Gribanova, Stanford