Chomsky’s derivation-by-phase theory as applied to judgments of the grammaticality of Japanese double accusative sentences

Four experiments explored the psychological reality of Chomsky’s derivation-by-phase theory. Speakers of Japanese judged the grammaticality of Japanese simple sentences involving two accusative arguments, each related with a single verb. In Experiment 1 one argument (a part-argument) constituted a part of the other argument (a whole-argument) and a part-argument either preceded or followed a whole-argument. Either the two arguments were concatenated or one argument was scrambled up to the sentence initial position. Judged grammaticality of the two types of sentences, concatenated and scrambled, was found comparably low, with no difference between them. Experiment 2 compared the two types of sentences with sentences including an adverbial phrase between the two arguments. The latter sentences were judged higher than the former for both the concatenated and the scrambled sentences. In Experiments 3 and 4, judged grammaticality increased as the distance between the two arguments was increased. Each of these findings could not be predicted by the derivation-by-phase theory. Implications of the findings for Chomsky’s mentalist position on speakers’ knowledge of language were discussed.