CNL Lunch Talks

Alison Austin, Ellen Lau, Clare Stroud

The Role of Structural Expectations in Detecting Structural Violations

Thursday April 15th, 2004, 12:30 PM, 3416 Marie Mount Hall

Electrophysiological studies of sentence processing have revealed a family of response components associated with different types of syntactic and semantic violations. Our studies seek to clarify the factors that distinguish early syntax-related ERP components (e.g. Early Left Anterior Negativity or ELAN, ~150-250ms) from later syntax-related components (e.g. P600).

In English, the ELAN has been reported only in response to sequences such as "...Max's of...". A number of different factors may be responsible for the earliness of violation-detection in this context: ill-formedness at the level of lexical categories, the violation of a strong expectation that the word following the possessor will be a noun, or the high-frequency of the function word "of".

In the first study, we used ellipsis to create contexts for the traditional ELAN comparison ('Max's proof of' vs. Max's of proof?) in which the expectation for a noun following the possessor was eliminated or weakened. If a strong structural expectation is crucial to the ELAN effect, we predict that we will see an ELAN only in the non-ellipsis conditions.

In the follow-up study, we once again manipulated the context surrounding 'of', this time using subcategorization to create grammatical/ungrammatical sentence pairs. This yields a comparison similar to the original Neville study, in which 'of' renders the sentence ungrammatical, but no longer violates a prediction.

A final comparison was done which tested whether the parser would register a mismatch to even grammatical sentences which violated the strongest prediction, e.g. an adjective after the possessive:

        The woman examined Max's very careful proof.

Based on our results, we will provide the beginning of an explanation which accounts for both the timeframe and limited distribution of the ELAN in terms of the ease of identifying ungrammaticality.