Abstract
Reinhart’s (2004, 2006) theory predicts that children should perform around chance on tasks involving reference-set computation due to their working memory limitations. The purpose of this article is to test her theory against acquisition data from a previously unexplored area of reference-set computation. We present an experiment designed to examine Mandarin-speaking children’s pronoun interpretation in contexts where the Avoid Pronoun Principle is operative. Since the computation of the Avoid Pronoun Principle requires reference-set computation, children are expected to show difficulty resolving pronouns in such contexts. This prediction was confirmed in our experiment, where otherwise linguistically competent 5-year-olds performed around chance when confronted with the need to compute the Avoid Pronoun Principle. Moreover, analysis of individual data suggests that their nonadultlike performance is due not to their lack of the relevant knowledge but to their lack of capacity to process the required computation.
Notes
1. 1For another bypassing strategy that is claimed to be available in other areas of reference-set computation, see Reinhart (Citation2004, Citation2006).
2. 2Several alternative analyses have been proposed to derive this principle from more general properties of language use (see, for example, Horn Citation1984; Levinson Citation1987; Y. Huang Citation1994), but the basic idea underlying all these proposals is that an optimal competitor is selected from a narrowly constructed set of alternative options.
3. 3It should be mentioned that the following example, where the pronoun is focused, is better than (5a).
(i)(Context: Zhangsan can sleep anytime anywhere with no difficulty.) Zhangsan1jinzhang-de(lian)ta1wanshangdoushuibuzhao. Zhangsannervous-DEevenhenightallcannot.sleep ‘Zhangsan was so nervous that even he couldn’t fall asleep at night.’
4. 4We treat the phonetically null subject as PRO rather than pro since its reference is obligatorily controlled by the matrix subject. Note, though, that whether the embedded subject is PRO or generalized Pro in the sense of C.-T. J. Huang (Citation1989) is not crucial, given that the Avoid Pronoun Principle applies to PRO and pro alike (Haegeman Citation1994).
5. 5The test sentences used in the diagnostic trials were V-de sentences with nonresultative interpretations (e.g., Xiaotuzi qie-de hen man ‘Little Rabbit cut it slowly’).
6. 6Needless to say, great care was taken in the experiment to ensure that the pronoun was not focused or rendered prominent in some other way.
7. 7Wilcoxon signed rank tests likewise indicated no significant differences between any pair of these conditions (all ps > .18).
8. 8This preference is independently confirmed by developmental studies such as C. Chomsky’s (Citation1969) experiment (Table 4.6, p. 108).