Published online by Cambridge University Press: 17 February 2009
This study investigated the degree to which cue validity, as estimated from textual analyses, predicts the actual strength of grammatical cues as they are used by speakers of Hebrew. An experiment was conducted to determine the differential strengths of the linguistic cues employed by Hebrew speakers when assigning the role of patient in sentences. Monolingual Hebrew-speaking subjects 4, 5, 7, and 9 years old, as well as adults, were tested using a sentence-picture verification task. Six cues were included in the study: word order, the accusative object marker, the reflexive noun phrase and three verbal derivations. By presenting subjects with sentences which set these cues in competition with one another, a measure of the strength of each cue was obtained. The results of a regression analysis revealed strong positive correlations between estimated cue validities and actual cue strengths for all but the youngest age groups. These results were interpreted as suggesting that cue validity is highly predictive of actual cue strengths. In addition, the strengths of the six cues varied as a function of the subject's age. Two additional factors were hypothesized to account for the performance of the older subjects: increased sensitivity to the reliability of cues and discourse-based constraints.
I would like to thank Ruth Berman, Brian MacWhinney, Janet McDonald and James McClelland for their helpful comments. I would also like to thank Lea Kalush for testing the children who participated in the study. This research was supported by NICHHD grant #HD 17790 to Brian MacWhinney.