Elicited production of relative clauses in children with Williams syndrome

Abstract

Relative clauses have been implicated alternately as a strength and a weakness in the language of people with Williams Syndrome (WS). To clarify the facts, an elicited production test was administered to 10 people with WS (age 10–16 years), 10 typically developing children (age 4–7 years), and 12 typically developing adults. Nearly every WS participant successfully produced both subject gap (SG) and object gap (OG) relative clauses, and produced relative clauses in both right-branching and centre-embedded positions. However, OG relative clauses were produced less frequently than SG relative clauses, and trials targeting OG relative clauses often elicited contextually inappropriate responses. Critically, these two patterns were also observed in typically developing children, though at attenuated levels. It is argued that by late childhood people with WS know the structure and the meaning of relative clauses, but qualitatively normal processing factors cause unusually severe interference in some subprocesses of sentence production.

Acknowledgements

This research was partially supported by the following grants: NSF SBR-9808585, March of Dimes FY98-0194 and FY99-0670 and NIH 1R03 HD43113. Thanks are due to Heather Jordan, Barney Pagani, and especially Nicole Kurz for help in conducting the study, to Barbara Landau for advice and support, to Gaby Hermon, Natalie Hsu, and especially Colin Phillips for fruitful discussions of this work, and to the children and families without whom this work would have been impossible.

This research was conducted by the author in partial fulfilment of a doctoral dissertation (awarded by Boston University; Zukowski, 2001). A sketch of some of the results of this study was provided in a recent review chapter (Zukowski, 2004). The pictures used in this study to elicit relative clauses were digitally altered versions of illustrations from Everyday Words in German (Litchfield, 1999). These illustrations were used by kind permission of Usborne Publishing Ltd.

Notes

¹Unfortunately it is difficult to prove this conclusively, because targeted structures could fail to be produced for other reasons too. For example, certain individuals or groups might respond differently to the contextual pressures that successfully elicit these structures from other participants. Fortunately, this is not an issue, given the results of the study reported here.

²The response was odd, but it successfully conveyed the picture change (e.g., target structure: The cat that the girl is chasing turned yellow, child's response: ‘That cat turned yellow because that girl is chasing it’. This participant's test session was curtailed midway. When she returned a week later to finish the test, she persisted with the alternative response that she had used the week before, and never produced another relative clause of any kind. Cases like this raise an important point. If this child had happened to produce her alternative response on the first trial rather than the 13th, she may never have produced a single relative clause, and this might have led to the incorrect conclusion that her grammar does not generate such structures at all. This highlights the difficulty of definitely proving lack of knowledge, even with elicited production. In cases where elicited production does fail to elicit a structure of interest, it is always wise to seek corroboration from other sources. Spontaneous speech could be examined, or sentence completion could be used as a method of precluding certain response possibilities.