Quantifier float and structure dependence in child Japanese

ABSTRACT

This study experimentally investigates whether Japanese-speaking children around the age of 4 conform to a structural constraint on Quantifier Float. After pointing out a potential confound in a previous study by Otsu (1994), I report the results of my own experiment, in which this confound in the experimental design is eliminated. The results of this new experiment confirm that the relevant structural constraint is indeed in the grammar of Japanese-speaking preschool children as suggested by Otsu, which in turn provides cross-linguistic support for previous findings regarding preschool children’s adherence to structure dependence.

Notes

¹ This is a simplified version of Miyagawa’s (1989) constraint, taken from Nakanishi (2008:290).

² See Sano (2007) for a related experimental study.

³ Basically the same consideration motivated the experimental study by Gualmini & Crain (2005). In their previous study that investigated the role of c-command in constraining children’s assignment of the conjunctive interpretation to sentences containing the disjunction operator or (Crain et al. 2002), English-speaking children were asked to interpret sentences like the following.

Table

As expected, children assigned the conjunctive interpretation of disjunction to sentences as in (ia) (in which negation c-commands the disjunction) but not to sentences like (ib) (which do not involve such a c-command relation). Gualmini & Crain (2005) observed, however, that these target sentences differ not only with respect to the structural relationship between negation and disjunction but also in the number of words that intervene between them: More specifically, negation and disjunction are closer in (ia) than they are in (ib). In light of this confound, Gualmini & Crain (2005) conducted a new experiment to demonstrate that children refrain from assigning the conjunctive interpretation to disjunction if this is merely preceded by negation, even if the two operators are very close to each other.

⁴ My choice of the predicate and the number of the animals used in the test sentences made it hard for me to adopt an act-out task, which Otsu (1994) employed in his experiment: I decided to use the predicate nokkatte-iru ‘be on top of,’ and it turned out to be quite difficult for children to act out sentences like ‘Three lions are on top of unspecified animals,’ which requires children to put three animals (e.g., lions) on the back of the two animals of a different kind (e.g., elephants).

⁵ In light of this potential ambiguity, one may wonder why this particle was chosen in my experiment, rather than other one-mora postpositions. Other candidates are -de ‘with (something),’ -e ‘to,’ -to ‘with (someone).’ The postpositions -de and -e typically take inanimate NPs in their complement position (as illustrated in (18) and (19)), and hence it is quite difficult to create a test sentence in which the antecedent of a floated quantifier is potentially ambiguous between the NP followed by a case marker and the NP followed by -de or -e: While the former is typically animate, the latter is not.

In a pilot stage of this experiment, I tested children with sentences involving -to ‘with (someone)’, as in (i).

Table

It turned out that children (as well as adults) had difficulty determining whether the numeral quantifier 3-biki refers to the number of rabbits or to the sum of rabbits and turtles. Whether the latter interpretation is in fact possible, and if yes, how this interpretation is derived from (i) are interesting issues that go beyond the scope of this study. Since it is apparent that the use of the postposition -to ‘with’ puts an extra burden on children, I decided to avoid the use of this postposition in the target sentences.

⁶ Sadakane & Koizumi (1995:12) observe that in the following sentence, which contains an intransitive verb expressing change of position, Quantifier Float over -ni is possible:

Table

Based on this observation, they argue that the particle -ni in this example is a case marker, not a postposition. Then, a question arises why the particle -ni behaves as a postposition in (29)–(31), despite the fact that the verbs in these examples and the verb in (i) are derived from the same verbal root (nor- ‘ride’). Even though I do not have a definite answer at this point, I suspect that the difference may be related to the distinction between distributive and nondistributive interpretations proposed by Kitagawa & Kuroda (1992:88–89): “the distributive construal necessarily implies the occurrence of multiple events while the non-distributive construal implies the occurrence of only a single event.” Building on this distinction, Ishii (1999) claims that apparent counterexamples to the mutual c-command requirement on floated numeral quantifiers permit a distributive reading but not a nondistributive reading and that the former interpretation involves an adverb-type numeral quantifier. Ishii’s proposal can account for the grammatical status of (i) even if we assume that -ni in this example is also a postposition as in (29)–(31): Since the example in (i) is forced to have a distributive reading, meaning that Kanta rode three horses one after another, the FNQ in this example is an adverb-type and hence does not need to satisfy the mutual c-command requirement.

I suspect that the stative predicate nokkatte-iru ‘be on top of’ in (28)–(31) is associated with a single event and hence excludes the possibility that the FNQ is an adverb-type, which in turn requires this FNQ to satisfy the structural constraint in (14). Whether this speculation is on the right track is an issue that needs more careful investigation, but I have to leave it for future research.

⁷ The correct responses for NP-ni NQ sentences as in (33) and for NQ-no-N sentences as in (34) were far above the chance level, by binomial test (p < .01, two tailed).

⁸ The 15 FNQ examples in the mother’s utterance contained the following classifiers: six instances of -tsu (general classifier for inanimates), four instances of -ko (general inanimate classifier for small objects), two instances of -ri (classifier for person), one instance of -dai (classifier for machines, vehicles, etc.), one instance of -hon (classifier for long, slender objects such as trees), and one instance of -too (classifier for large animals).

⁹ I would like to thank an anonymous reviewer for suggesting this type of alternative account to me.

Additional information

Funding

This research was supported in part by a Grant-in-Aid for Scientific Research (C) from the Japan Society for the Promotion of Science (Grant Number 25370550).