2.2. Materials

In this study, we changed the word order of complex sentences in Japanese, including a quantifier, to manipulate the presence and absence of the syntactic integration of discontinuous constituents, the violation of a syntactic island constraint, and the syntactic prediction in syntactic integration. We constructed sentences of six phrases in Japanese, in which the order of the quantifier and its head noun was changed in a two-by-three way, that is, two precedence relationships between a quantifier and its head noun (Quantifier First and Head-noun First) and three distances between the two (Adjacent, In Syntactic Island, and Distant), as in Table 1.

Examples of the experimental sentences.

The (possible) propositional meanings of the six sentences were identical to Mr/Ms. Ohta took six pictures (while he/she was) walking in the park. The underlined words indicate quantifiers and their head nouns, and the square brackets ([ ]) indicate adverbial phrases as possible syntactic islands.

In the current study, we did not assume the a priori effect of a syntactic island violation for Japanese because the syntactic island effect in Japanese could be relatively weak compared to the effect in English. Sprouse et al. (2012) experimentally examined the syntactic island effects in various English constructions. When creating experimental sentences, Sprouse et al. (2012) managed to keep the propositional meaning of a subordinate clause unchanged and manipulated the presence or absence of a construction that could function as a syntactic island. With this manipulation, Sprouse et al. (2012) succeeded in observing the effect of the discontinuous dependency and that of the syntactic island independently. Tokimoto (2019) applied the analysis method in Sprouse et al. (2012) to Japanese sentences that directly corresponded to the experimental sentences in Sprouse et al. (2012), and Tokimoto (2009) demonstrated that the island effect in Japanese was much weaker than that in English. Tokimoto (2009) examined the island effects in Japanese by manipulating the long-distance scrambling of a subordinate object, whereas in the current study, the presence and the absence of a possible island effect were manipulated by a quantifier floating. We were thus careful in predicting the island effect.

The theoretical interests for each word order are summarized below. The dependency relationship between a quantifier and its head noun is expected to be constructed during the fifth phrase (P5). In (c, Quantifier First, Distant) and (f, Head-noun First, Distant) in Table 1, a subordinate clause intervenes between the quantifier and its head noun, and therefore, a syntactic integration process of the discontinuous quantifier and head noun is expected at P5. For the examination of the effect of syntactic integration, (a, Quantifier First, Adjacent) and (d, Head-noun First, Adjacent) are the controls for (c) and (f), respectively. In (c), the following head noun is predicted at the input of a quantifier, whereas in (f), a quantifier is not predicted at the input of a (head) noun.

Roku-mai (six) in (b, Quantifier First, In Syntactic Island) and shashin (picture) in (e, Head-noun First, In Syntactic Island) have to establish the dependency relationship with shashin in (b) and roku-mai in (e), respectively. However, one of the two phrases is placed in a possible syntactic island of an adverbial clause in (b) and (e); therefore, the dependency between the two phrases can be difficult to construct.

Furthermore, in (b) and (e), the violation of a syntactic island constraint can be detected at the subordinate verb sanpo-shitsutsu (a walk-taking), at which the verb and teien-o (park-acc) can construct an adverbial clause. That is, the parser can recognize that a quantifier or a (head) noun is placed in an adverbial clause at the end of the clause even before the parser has received the corresponding head noun or the quantifier. This situation is similar to the example by McKinnon and Osterhout (1996) in that the presence of a wh-island can be recognized at when placed at the beginning of the subordinate clause.

The quantifiers in the experimental sentences take inanimate nouns as their head nouns. We constructed thirty sentences for each word order: a total of one hundred and eighty sentences. The fifteen subordinate verbs were suffixed by the connective particle -nagara, and the others were suffixed by the particle -tsutsu, both of which mean while. Thirty ungrammatical controls were included in the main session, and in total, two hundred and ten sentences were divided into 3 blocks. The ungrammatical controls were deviant in their argument structures. Two of them are shown in (14).