This paper provides basic information on the KY Corpus, a frequently used collection of natural but highly structured conversation data from 90 Oral Proficiency Interviews (OPI)，consisting of three groups of speakers of Japanese as a second language with native language backgrounds in English, Chinese and Korean (30 in each group), and representing four different proficiency levels: Superior (5 interviews), Advanced (10), Intermediate (10) and Novice (5). It also discusses the strengths and weaknesses of the corpus, referring to the numerous investigations in which it has been used either as a means of verifying theoretical claims regarding second language acquisition or as the main object of research on OPI. In anticipation of the increasing use of computer programs such as CHILDES and ChaSen, it is hoped that researchers will pay due attention to the nature of the data that they analyze.

　This study examines the actual state of use of the final particle wa, based on the natural language data provided in Josei no kotoba: shokuba-hen (Women's Speech: The Workplace). Building on previous research that has given a quantitative overview of the data, this study focuses on a qualitative analysis. The following three relative characteristics were observed:(1)Final particle wa is predominately used in informal settings with listeners of the same age or younger with whom the speaker has frequent contact. Accordingly wa is never found co-occurring with the polite forms desu and -masu. (2) Younger speakers have a greater tendency to use final particle wa without co-occurring final particles ne and yo, and is often uttered with an extended intonation. (3) Among younger speakers, the use of wa is particularly common with certain topics. While the overall frequency of use of wa among younger speakers is certainly not great, they do use it as one means affective expression, which suggests that it would be premature to conclude that it is in the process of dying out.

　Analyzing compositions written by intermediate-level Chinese learners, this study investigates both quantitatively and qualitatively the acquisition of adjectival collocations and examines LI transfer effects. Specifically, the study focuses on collocations involving nouns and i-adjectives and nouns and na-adjectives. Error analysis revealed that (1) almost 20% of the composition collocations were errors and (2) while learners tend to use more basic attributive i-adjectives, na-adjectives are more frequent at advanced levels. Based on these findings, the paper discusses the following pedagogical implications for improving adjectival usage: (1) applying Chinese-Japanese contrastive analysis to adjective instruction, (2) extracting examples for polysemous adjectives from corpora, (3) emphasizing typical restrictions on adjective usages, and (4) exploiting positive transfer effects from Chinese for the effective instruction of na-adjective usages.

　This paper attempts to clarify the differences of "Writer/Reader visibility" (overt presence of participants) between essays written by Chinese-speaking learners of Japanese (CN) and those by native speakers of Japanese (JP), by comparing interrogative sentences based on the theory of Dependence on Hearer's Information. We chose 43 opinion essays written by CN and 44 by JP from a corpus collected by the National Institute for Japanese Language. The results of the analysis are as follows:

1)　CN tend to use interrogative sentences "dependent on Hearer's Information" more often than JP do. Especially, CN often use polite interrogative forms (i.e. desu ka / masu ka), while JP never do.

2)　CN use interrogative sentences in the corpus with the function of "asking questions", while JP don't.

3)　Both CN and JP use interrogative sentences with the function of "raising issues", but CN have not acquired the appropriate forms, that is, interrogative sentences "not dependent on Hearers Information" (i.e. daroo ka / desyoo ka).

　These findings suggest that interrogative sentences in CN writing show heavier use of functions that interact with and appeal to the reader than those in JP writing, and that CN writers of Japanese compositions accordingly use more features of Writer/Reader visibility than JP do.

　This paper investigates beliefs of 415 non-native Japanese teachers who teach Japanese outside Japan, based on a factor analysis of their belief characteristics. The results show two significant factors which can be categorized as accuracy-oriented (seikakusa-shikō) and expressiveness-oriented (yutakasa-shikō). Teachers with the former factor tend to take language structures and accuracy as more important than other entities, while those with the latter factor place priorities on culture and motivation. Further, the factor scores show which attributes have an impact on these factors: the gender of the teachers, the geographical region, their Japanese ability, the academic grades of the students are found to have explicit impacts on both factors. The experience of learning in Japan is found to have an explicit impact on the expressiveness-oriented factor only. The study finds the complex of attributes associated with each factor: while the region of teachers' residence shows strong impact on the accuracy-oriented factor, the language education policy might have influenced the non-native Japanese teachers' beliefs. With these results, the study might be able to contribute to understanding non-native Japanese teachers' beliefs.

　Making a compliment is generally considered to be one of the positive politeness strategies, but may result in unintentionally offending the addressee. The risk seems to be higher in intercultural communication because appropriate use of a compliment differs from culture to culture. Making an appropriate compliment is an important part of pragmatic competence, and it is meaningful to find the factors that make a favorably-intended compliment offensive.

　This paper analyses evaluative utterances including compliments or criticisms by three standards: expected effect (friendly / offensive); intended evaluation (positive / negative); and literal evaluation (positive / negative). The results show that utterances have offensive effects when they make 1) explicit reference to the superiority or sense of self-praise of the addressee; or 2) reference to a topic that the addressee does not want to be closely observed. Preferring an implicit compliment over an explicit one and disguising a positive evaluation as a negative one, both of which often puzzle non-native speakers, are some of the strategies to avoid possible offensive effects of a compliment.