Corpus counts fo NP sequences in Japanese
Ryu Iida, Edoson T. Miyamoto · NAIST Digital Library (Nara Institute of Science and Technology) · 2002
It is often assumed that sentences are processed incrementally in the sense that words are associated to each other immediately as they are read o a sentence.There is no delay as people do not wait for the end of the sentence in order to start understanding the meaning conveyed.However, the incremental processing of sentences in head-nal languages such as Japanese is counter-intuitive because it requires words to be associated to each other even when no direct relation can be established between them (see Miyamoto, 2002, and references therein for discussion).Compare the following fragments in English and in Japanese respectively.(1) a. Mary gave . . . b. Mary-ga John-niMary-nom John-dat. . .The two fragments exemplify an often-encountered dierence between sentences in the two languages: In English, the verb is the second constituent of the clause; in Japanese, a sequence of NPs occursbefore any predicate is encountered.Consider the associations that readers have to perform as they reach the second word in each fragment.In the English fragment in (1a), Mary can be associated as the subject of the verb gave.In the Japanese fragment, in contrast, there is no predicate-argument relation that can be established between the two NPs.Nevertheless, the case markers suggest that the nominative NP Mary is likely to be the agent, and the dative NP John the goal of an upcoming predicate.The question then arises as to how Japanese readers associate NPs within clausal structures before any predicate is encountered.In other words, when a NP sequence is detected in the input string, how many predicates should readers expect and what would betheir preferences in terms of which NPs should beassociated with which predicates?For example, a sequence of two nominative NPs (NP-ga NP-ga) allows two types of continuations.In one type, one single predicate suÆces and the two NPs are co-arguments associated to this single predicate.Alternatively, two predicates may be expected, with each NP-ga being associated to a dierent predicate.Several psycholinguistics proposals (e.g.minimal attachment, Frazier, 1987; simplicity, Gorrell, 1995; inter alia) predict that the single predicate interpretation should be preferred.However, various language-speci c factors suggest that a nominative NP occurring after a subject NP tends to indicate the beginning of a new clause (Miyamoto, 2002, for experimental evidence; also Nagai, 1995, for related discussion).Two corpus counts were conducted in order to determine the frequency with which nominative marked NPs in mid-sentence coincide with the beginning of embedded clauses 1 Technical Report No. 2002015