抄録
In this paper, we will show a surprising parallelism of behaviour in spite of the difference in meaning between bare nouns in Japanese and definite noun phrases in English, and try to explain it in terms of discourse processing. We assume that linguistic input is processed in relation with the discourse resources consisting of three distinct databases: general knowledge, utterance situation and discourse memory. Discourse memory is further divided into three subdomains: linguistic data memory, linguistic understanding memory, and long-term discourse memory. For each of these five domains, some means must be available for identifying elements there. Pronouns can only search in the linguistic data memory and demonstratives, in the utterance situation or the linguistic data memory. Definites in English and bare nouns in Japanese usually express roles with parameters left unspecified. They yield values when the parameters are specified, but they don't tell how to identify the appropriate parameters. This underdetermination is exactly what is needed for them to be able to operate a search in any of the domains within the discourse resources. If any language needs a means to signal this kind of default search and if only bare nouns in Japanese and definites in English can fulfil this, then they will. We thus hypothesize that the parallelism of behaviour of definites in English and bare nouns in Japanese comes from this kind of general principle of language.
