CNL Lunch Talks

Andrea Zukowski & Jaiva Larsen

The Production of Sentences

That We Fill Their Gaps

Thursday November 20th 2003, 12:30 PM, 3416 Marie Mount Hall

Sometimes the way we begin a sentence precludes the possibility of finishing it in a grammatically correct way, given the particular message that we have chosen to convey (we 'talk ourselves into a corner'). We (unintentionally) discovered an experimental context that induces some unimpaired adults to produce ungrammatical sentences whose ungrammaticality is not apparent until the very end of the sentence. The errors involve filling an object gap in a relative clause with a body part, as in "The woman who the caterpillar is falling on her head." We have evidence to suggest that adults who make such errors do not do so because they think filled gaps are grammatical. We believe that understanding the cause of 'implementation errors' like this will help us to understand other production errors that we have observed in people with Williams syndrome--a neurodevelopmental disorder that results in mild to moderate mental retardation, but fluent language that, although not perfect, is impressive in its syntactic complexity.

In the study that we present today, we investigated the question of whether filled gap errors like the one given above are predictable on the basis of two cognitive variables (inhibition and working memory) that are likely to be of use when you find yourself in the middle of a sentence that is destined to go wrong if it continues on its present course. Our subjects were 29 unimpaired adults. For each subject, we administered measures of inhibition (Logan) and working memory (backward digit span), elicited relative clauses of two kinds (subject gap and object gap, both with and without 'body part contexts'), and administered a grammaticality judgment task. One third of the subjects never attempted to produce an object gap relative clause, so they could not have filled the object gap site. Of the remaining 20 subjects, half of them produced the filled gap error. Also, to our surprise, 6 of the 29 subjects produced 'mapping errors' one or more times. These are responses that are grammatically well-formed, but whose meaning does not convey the message that was intended (e.g. Experimenter asks: "Which man turned purple?," subject responds: 'The eagle who's landing on the man's shoulder'). Ironically, these are precisely the errors that we are seeking to understand in people with Williams syndrome. We examine the data to determine whether people who fill object gaps or people who produce mapping errors differ from those who do not in either inhibition or working memory abilities.