Segmentation and selection of appropriate Chinese characters in writing place names in Japanese.

This paper explores the relation between an unknown place name written in hiragana (a Japanese syllabary) and its corresponding written representation in kanji (Chinese characters). We propose three principles as those operating in the selection of the appropriate Chinese characters in writing unknown place names. The three principles are concerned with the combination of on and kun readings (zyuubako-yomi), the number of segmentations, and the bimoraicity characteristics of kanji chosen. We performed two experiments to test the principles; the results supported our hypotheses. These results have some implications for the structure of the Japanese mental lexicon, for the processing load in the use of Chinese characters, and for Japanese prosody and morphology.

Processing discontinuous words: on the interface between lexical and syntactic processing.

How are discontinuous words processed? Are they identified in the lexicon or in the syntax? Schreuder (1990) proposes the existence of morphological integration nodes (MI nodes) to account for the representation of complex verbs with separable prefixes in Dutch. We tested the MI model during sentence processing in Dutch, using an ungrammaticality judgment task. The results supported the predictions of the Schreuder model, and also provide evidence for distinct lexical/morphological and syntactic processing subsystems, each driven by the information resources and tasks relevant to its own representational vocabulary. It is argued that no special principles are needed to govern the interaction of lexical/morphological and syntactic processing, even for the identification of discontinuous words; this follows automatically from independently required characterizations of the subsystems themselves. We also examine the principles underlying the MI model in an attempt to extend the model to a wider array of constructions and languages. It is hypothesized that frequently encountered linguistic expressions are represented in the lexicon. In the basic case, they are represented as access nodes, if they may stand alone, and as MI nodes, if their constituents are already represented by access nodes. Unlike the original MI model, no further stipulation is needed concerning the existence or inhibition properties of MI nodes, assuming that candidate lexical/morphological hypotheses are appended to whatever portion(s) of the input string they are hypotheses about.