1. DeKeyser reports that the fuzzy rules were learned better by the implicit-inductive subjects than by the explicit-deductive subjects, but he also notes that the results regarding the fuzzy rules should be treated with caution, due to the small data set and the lack of statistical testing.

2. "Comprehensible input" is defined by Krashen as input containing i+1, where i is the acquirer's current level and i+1 is structure a bit beyond the current level.

3. Only the processing group was informed that language learners often misinterpret NVN sequences (1993a, p. 232); the processing group practiced OVS sentences but the traditional group did not (1993b, p. 53); and so forth.

4. The processing instruction adopted the "meaning-form connection" approach (Terrell, 1991), which encourages the learners to understand the concepts and meanings underlying the grammatical forms, but the traditional instruction followed a more mechanical, paradigmatic presentation of grammatical forms.

5. Cadierno (1995) extended the study using Spanish past tense verb morphology and obtained results similar to those in VanPatten and Cadierno (1993a, 1993b).

6. According to Swain, "comprehensible output" means "output that extends the linguistic repertoire of the learner as he or she attempts to create precisely and appropriately the meaning desired," (Swain, 1985, p. 252).

7. Krashen discusses the relative complexity of a target structure in terms of formal and functional/conceptual complexity (1987, pp. 97-98), so the term "complexity" in the present paper includes both formal and functional/conceptual complexity of a structure.

8. The description of the Japanese honorific system in this paper is based on Clancy's (1985) study, "honorifics and late acquisition of L1 Japanese", the in-group and out-group concepts are from Jorden's textbook (1987), the honorific usage is provided by Simon's grammar notes (1987) and Mizutani and Mizutani's textbook (1977), and by discussions with Kyoko Suda (Instructor of Japanese, University of San Francisco).

9. There is another regular, respectful verb form, "V-(r)areru," but this study focuses on the form "o + verbal stem + ni narimasu." The romanization used in this paper is an adaptation of the Shin-kunrei-shiki 'New official system' (Jorden, 1987).

10. The concepts of in-group and out-group also shift depending on context. For example, when a speaker belongs to Company A and talks to a person who belongs to Company B, the speaker uses a humble form to describe not only his or her own action but also the actions of other members in Company A (including the speaker's superior) because all members in Company A are now members of the speaker's in-group when speaking to a person in a different company (out-group). Since the students in this study were still at an elementary level and were introduced to Japanese honorifics for the first time, the situations in which the students were instructed to use honorifics in the program were mostly limited to a school situation (in which a student talks to his or her teacher or talks about the teacher with other people), or a company situation (in which an office worker talks to his or her superior or talks about the superior with his or her colleague). Examples of out-group/in-group shift were not included in the program.

11. The student who obtained the highest score and the student who obtained the second highest score on the mid-term exam were paired and were randomly assigned to one of two groups. The same pairing system was applied until the last two students (the students with the two lowest scores) were paired and randomly assigned to one of two groups.

12. The two-sample dependent t-test (the paired two-sample t-test) was applied to examine whether there is a significant difference between the input-focused group and the output-focused group in their scores on the mid-term exam. See Glasnapp and Poggio (1985) for a description of the t-test for correlated samples. The mean score of the input-focused group on the mid-term exam was 86.9 (SD = 11.6) and that of the output-focused group was 88.2 (SD = 9.3).

13. One might argue that since Korean has its own honorific system, the student with a Korean background may have had an advantage in learning Japanese honorifics. However, the Korean student is included in the input-focused group in this study, so this argument strengthens the result, since the output-group actually performed better than the input-group. A statistical analysis was also performed to check the results when the Korean student (in the input-focused group) and the student paired with the Korean student (in the output-focused group) were excluded from the study. The results were still significant, favoring the output-focused group.

14. In order to reduce the lexical load, the computer programs employed only vocabulary that had been introduced before (except for four nouns such as katyoo, 'section chief', butyoo, 'department chief', etc., and two verbs, yomimasu, 'read' and hanasimasu, 'talk', which were new).

15. The input-focused program employs the multiple choice format because it is typically used for comprehension exercises. However, multiple choice exercises may not necessarily capture the pedagogical benefits of all types of comprehension exercises. Some other question formats (such as asking learners to translate the target language into their first language by themselves) will be included in a future study.

16. The input-focused program stores a feedback message corresponding to each response to a multiple-choice question, so when the student selects a number, the corresponding message is provided.

17. The Japanese word processor works in the following way: the students type roman letters on the alphabetical keyboard and these roman letters are converted to katakana, hiragana, and kanji. (Katakana and hiragana represent Japanese syllables, so they are phonetic descriptions. Katakana are used to write words borrowed from English or other foreign languages. Hiragana are used mainly to write particles and grammatical inflections in verbs, adjectives, and adverbs. Kanji represent both sound and meaning. Japanese sentences are normally written with a mixture of hiragana, katakana, and kanji.) For homonymous words with the same romanization, the students obtain a list of kanji for the word and are asked to select the kanji that he/she thinks are appropriate.

18. The output-focused program analyzes the learner's response using a pattern matching technique: the program stores correct words, phrases, sentences, and anticipated errors for each question, and checks whether these items are found in the learner's response. (No parsing technique is involved in this study.) The program also stores feedback messages corresponding to correct answers and anticipated errors, so if the learner's response matches a correct answer or an anticipated error, a corresponding error message is provided. Anticipated errors include missing words/particles, incorrect/incomplete verbal forms, and incorrect Japanese writing systems.

19. Giving more exercises to the input-group also strengthens the result if the output-group performs better than the input-group, which is indeed the case.

20. Among 137 exercises, 41 exercises were of type 1, 45 of type 2, 35 of type 3, 8 of type 4, and 8 of type 5.

21. Among 130 additional comprehension exercises, 66 exercises were of type 1, 22 were for auditory comprehension of spoken honorific verbs (a combination of type 1 and type 5), 41 were for auditory comprehension of spoken honorific sentences (a combination of type 3 and type 5), and one was of type 4. The additional exercises for the input-focused group were intended to review comprehension of honorific verbal forms and to practice auditory comprehension of honorific sentences. Consequently, type 1, type 3, and type 5 tasks were mostly added for the input-focused group.

22. The first computer session focused on the irregular respectful forms and the irregular humble forms, the second computer session included the regular respectful form "o + verbal stem + ni narimasu", the third computer session introduced the regular humble form "o + verbal stem + simasu", and the fourth session reviewed all kinds of honorific sentences practiced in the previous sessions.

23. Different numbers of questions were drawn from the various question types on the achievement test. It should be kept in mind, however, that questions of different types could vary in the time and amount of work required to complete them. In particular, time constraints limited the number of type 4 questions that could be administered. Development of a more principled measure is an important topic for future study.

24. Type 3 production scores exhibit the largest difference between the two groups on the achievement test, and the retention production test included only type 3 tasks. The t-score might have been lower if other types of tasks were included in the retention test.

25. When multiple t-tests are performed on a series of null hypotheses, the probability of committing at least one type I error (i.e., the probability of mistakenly rejecting the null hypothesis) may be higher than the significance level employed in each of the tests. To achieve a significance level of alpha for the conjunction of the null hypotheses, it suffices to divide the significance level of each of the individual tests by the number of tests performed (Galambos & Simonelli 1996). Accordingly, suppose we select a significance level of 0.05 for the present study as a whole and perform t-tests for the mid-term exam, the comprehension achievement test, the production achievement test, the comprehension retention test, the production retention test, and the oral conversation test. In this case, the significance level of each test becomes 0.008 (0.05 divided by 6). Only the production achievement test is significant at the 0.008 level. The tests that fail at this corrected significance level may nonetheless be regarded as heuristic indications for future study with a larger sample size.

26. P = 0.179 for item 3, p = 0.111 for item 9, p = 0.194 for item 18, p = 0.410 for item 19, p = 0.277 for item 20, and p = 0.334 for item 21.

27. The new programs are BANZAI: NOMINAL MODIFIERS (consisting of two lessons on Japanese nominal modifiers), BANZAI: CONDITIONALS (consisting of two lessons on Japanese conditional expressions), BANZAI: CAUSATIVES (consisting of two lessons on Japanese causative expressions), BANZAI: SENTENCE MODIFIERS (consisting of two lessons on Japanese sentence modifiers), BANZAI: GERUNDS (consisting of three lessons of Japanese expressions with verbal gerunds), BANZAI: VERBAL DIRECT STYLE (consisting of three lessons of Japanese expressions with direct-style verbs), and BANZAI: GIVING AND RECEIVING (consisting of three lessons of Japanese giving and receiving expressions). Anyone who is interested in the programs may request them from the author (nagatan@usfca.edu).