CNL Lunch

Carol Whitney

How a proposed neural parse representation can account for some syntactic complexity effects

Thursday October 18th 2001, 12:30pm, 3416 Marie Mount Hall

I will address the question of how the syntactic structure of a sentence could be represented by the brain. The proposed encoding is neurobiologically plausible, and explains some linguistic and psycholinguistic phenomena. The encoding is comprised of two levels. A temporal representation splits the incoming sentence into phrases; this encoding is converted into a non-temporal representation of the hierarchical structure of the sentence, in terms of thematic roles. I will sketch how these encodings could be implemented with neurons, and demonstrate how they could be used to process sentences.

I will conclude by showing that the proposed processing can account for the following phenomena:

(1) Most languages have a consistent ordering of heads within phrases.

(2) Multiple cross-serial embeddings in Dutch are easier to understand than center-embedded nestings in German.

(3) In English, a doubly center-embedded relative clause is too complex to understand; it seems felicitous to drop the middle verb, or to have the innermost subject be a first- or second-person pronoun.

(4) In Japanese, a sentential complement within a sentential complement is acceptable, but an adverbial clause within a sentential complement is too complex.