Pragmatic Comprehension in Japanese as a Foreign Language

Pragmatic Comprehension in Japanese as a Foreign Language NAOKO TAGUCHI Carnegie Mellon University Modern Languages BH160 CMU Pittsburgh, PA 15232 Email: [email protected] This study developed an original instrument that measures pragmatic comprehension in Japanese as a foreign language (JFL). It examined the ability to comprehend implied meaning encoded in conventional and nonconventional features and the effect of proficiency on comprehension. There were 63 college students of Japanese at 2 proficiency levels who completed a listening test measuring their ability to comprehend 3 types of implied meaning: indirect refusals, conventional indirect opinions, and nonconventional indirect opinions. Comprehension was analyzed for accuracy (scores on a multiple-choice measure) and comprehension speed (average time taken to answer items correctly). There was a significant effect of item type on accuracy but not on comprehension speed. A proficiency effect was observed on accuracy but not on comprehension speed. Analyses of error data and introspective verbal reports revealed the nature of comprehension difficulty among JFL learners.

THIS RESEARCH EXAMINED AN ASPECT OF pragmatic competence that is rather underrepresented in the literature of interlanguage pragmatics: pragmatic comprehension. Pragmatic comprehension involves the inferential process of understanding what a speaker intends to accomplish by making utterances (Thomas, 1995; Verschueren, 1999). Although a growing body of second and foreign language (L2/FL) research has examined learners’ abilities to understand speakers’ implied intentions, most previous studies have been confined to English as a second or foreign language. The present study aimed to fill the gap by developing an original listening instrument that measures pragmatic comprehension in Japanese as a foreign language (JFL). The instrument incorporated conventional linguistic features of indirect communication in Japanese that have been established in the previous literature. The learners’ pragmatic comprehension, operationalized as the ability to understand conventional and nonconventional implied meaning, was compared across different proficiency levels. The Modern Language Journal, 92, iv, (2008) 0026-7902/08/558–576 $1.50/0  C 2008 The Modern Language Journal

The proficiency effect on two attributes was examined: accuracy and speed of comprehension. BACKGROUND Pragmatic Comprehension in an L2/FL Previous studies have examined pragmatic comprehension under a variety of labels: comprehension of (a) speech acts, (b) conversational implicatures, (c) routines, (d) indirect utterances, and (e) implied speakers’ intentions. These studies addressed a range of issues, including the proficiency effect on pragmatic comprehension (Cook & Liddicoat, 2002; Garcia, 2004; Koike, 1996; R¨over, 2005; Taguchi, 2005, 2007), the development of pragmatic comprehension (Bouton, 1992, 1994; Taguchi, 2007, 2008a), the effect of learning context on pragmatic comprehension (R¨over; Taguchi, 2008a), the effect of different implicature types on comprehension (Carrell, 1984; Cook & Liddicoat; Garcia; Koike; Taguchi, 2002, 2005, 2007), the comprehension speed of indirect utterances (Holtgraves, 2007; Takahashi & Roitblat, 1994; Taguchi, 2005, 2007, 2008a,b), and the relationship between pragmatic comprehension and production (R¨over).

Naoko Taguchi Previous research revealed a complex interplay among types of implied meaning, general L2/FL proficiency, and pragmatic comprehension. Some implied meanings are easy to understand, whereas others are difficult, and proficiency affects this comprehension. Cook and Liddicoat (2002), for instance, examined comprehension of indirect L2 requests. In their study, learners of English as a second language (ESL) responded to a written test that measured their ability to comprehend three types of request expressions: direct requests (e.g., “Please” + verb), conventional indirect requests (e.g., “Can you” + verb), and nonconventional indirect requests (e.g., hinting). The results showed that more proficient learners were able to understand direct and conventional direct requests better than less proficient learners. Similar effects were observed in Bouton’s (1992, 1994) studies, which compared ESL comprehension over six types of implicatures: relevance-based implicatures, “Pope” implicatures (e.g., showing agreement by saying, “Is the Pope Catholic?”), indirect criticism, sequencebased implicatures, irony, and scalar implicatures. Over 4.5 years, the ESL learners’ comprehension of implicatures improved, but they still had problems with Pope implicatures. The learners’ first language (L1) backgrounds also affected comprehension. Korean, Japanese, and Chinese speakers showed significantly poorer comprehension than did the German, Spanish, and Portuguese L1 groups. These findings suggest the effects of conventionality on L2 pragmatic comprehension. As Cook and Liddicoat (2002) showed in their study, conventional request forms were easier to comprehend than nonconventional, hinting expressions, probably because they were easily associated with their functions. These associations were not straightforward for the nonconventional requests, resulting in comprehension errors. Bouton’s (1992, 1994) findings suggest that conventionality facilitates comprehension only when learners are familiar with the conventions. The Pope questions remained difficult for the learners probably because the conventions were culture-specific and foreign to them. When learners are aware of the norms of interaction and language use, the meaning of conventional implicatures becomes transparent, as long as they have sufficient linguistic skills to comprehend the utterances. Knowledge of conventions does not develop naturally over time, as shown in Bouton’s (1992, 1994) studies. These conventions need to be taught explicitly or experienced frequently

559 in the L2/FL context (Bouton, 1999; Kubota, 1995). These findings corroborate pragmatic theories that explain underlying cognitive mechanisms in the process of comprehending nonliteral meaning. Relevance theory (Sperber & Wilson, 1991, 1995) claims that implicatures differ in their degree of strength; some are strongly conveyed, whereas others are weakly understood due to the number of contextual cues that need to be processed. The greater the number of cues to be processed, the more extensive the listener’s search for meaning becomes, which results in a greater processing load. In contrast, when implicatures contain conventional lexical and syntactic features, or formulaic expressions, or they are embedded in predictable discourse patterns, listeners do not have to rely on extensive analysis of contextual cues to understand the meaning behind the implicatures. The facilitative effect of conventionality in comprehension has been documented in psycholinguistic studies that have examined the speed involved in literal and nonliteral comprehension. The studies have typically used response times as a measure of comprehension speed. Comprehension is faster when the preceding proposition is immediately accessible. When the proposition is not salient, listeners need to construct some sort of bridging structure, resulting in a longer response time (e.g., Bezuidenhout & Cutting 2002; Gibbs, 1981, 1983; Hamblin & Gibbs, 2003). Conventionality enhances comprehension speed because, due to the routinized associative connections in long-term memory, conventional utterances do not require extensive analytical procedures (Rumelhart, 1993). Several L1 studies have shown that conventional utterances take a shorter time to process than nonconventional utterances (e.g., Hamblin & Gibbs). In L2/FL, my study (Taguchi, 2005) documented the facilitative effect of conventionality in comprehension. The study measured the ability to comprehend more and less conventional implicatures among English as a foreign language (EFL) learners. More conventional implicatures included common refusal routines (giving a reason for refusing) and conventional request expressions (e.g., “Do you mind” + verb). Less conventional implicatures consisted of indirect opinions that did not have specific linguistic expressions or language use patterns. For example, expressing a negative opinion of a gift with the statement “The wrapping paper was nice” illustrates the wide variety of linguistic options for

560 expressing opinions (e.g., endless qualifications of liking or disliking). Comprehension was assessed for both accuracy and comprehension speed (response times). The results revealed shorter response times for the more conventional implicatures. There was a significant proficiency effect on accuracy but not on comprehension speed. Accuracy and comprehension speed bore no significant relationship. My longitudinal studies (Taguchi, 2007, 2008a, b) also revealed similar effects of implicature types on comprehension speed among EFL and ESL learners. Comprehension was faster and more accurate for indirect refusals than for indirect opinions. The learners showed a significant improvement in both accuracy and comprehension speed over time, but the degree of gain was greater for the comprehension of indirect refusals than for indirect opinions. The ESL learners showed a more profound gain in comprehension speed than in accuracy over time, whereas the pattern was reversed for the EFL learners, who gained little in comprehension speed over time. To summarize, previous research has established a relatively well-formed analysis of the nature of pragmatic comprehension in an L2/FL. The degree of indirectness is closely related to the amount of processing effort involved in comprehension. The processing effort is affected by linguistic and nonlinguistic features, such as features of conventions, cultural and linguistic knowledge, and frequency of the utterances. The processing effort manifests in differences in learners’ accuracy and speed of comprehension (i.e., response times) over a range of implicature types, and these differences indicate a hierarchy of difficulty among implied meaning types. Learners’ comprehension also interacts with learner-specific factors, including proficiency, length of residence, and context of learning. These generalizations drawn from the previous research literature remain largely in the area of L2/FL English. With the exception of Koike’s (1996) study, which looked at the comprehension of indirect suggestions by learners of Spanish, the target language has been almost exclusively English, providing little crosslinguistic generalizability of the findings. Hence, future research should investigate pragmatic comprehension across different target languages. Given the crosslinguistic variations established in the speech act literature, it could prove fruitful for future studies to analyze features of indirect communication in a specific language and incorporate them into an instrument that measures comprehension of indirectness in an L2/FL. The present study is an effort

The Modern Language Journal 92 (2008) toward this goal in the context of Japanese as an L2/FL. Pragmatic Competence in Japanese as an L2/FL Pragmatic competence in L2/FL Japanese has been examined from a variety of perspectives. The richness of the literature is symbolic of the fact that Japanese linguistics has taiguu hyoogen as a distinct area of study. Taiguu hyoogen is defined as a category of linguistic expressions that affects and is affected by social relations in a society (Mizutani, 1985). It includes elaborate categorizations of honorifics, speech styles, and expressions of reservation and intimacy, which are considered to encode social organization directly into the Japanese language. Previous research has examined linguistic usage related to those categories to make inferences about pragmatic competence in L2/FL Japanese. Abundant research has examined learners’ use of politeness features—in particular, honorifics, speech styles, and address terms (Cook, 2001; Marriot, 1995; Miyaoka & Tamaoka, 2001; Takenoya, 2003). For instance, the Japanese language is known for its elaborate system of address terms that range over three second person pronouns (anata, kimi, and omae) and five honorific titles (sama, san, sensei, kun, and chan), which differ according to the gender, age, and social positions of the interlocutors, as well as settings of communication. Takenoya’s crosssectional study revealed that even advanced-level learners had difficulty using the address term system appropriately. They mastered second person pronouns and honorific titles but could not identify the speaker when the address terms were embedded in conversations. Research in the area of speech acts has also revealed challenges that learners of Japanese face when performing illocutionary functions (e.g., Furukawa, 2001; Ikoma & Shimura, 1993; Jones, 1998; Kashiwazaki, 1993; Mizuno, 1996; Saito & Beecken, 1997; Sameshima, 1998; Tateyama, 2001). Ikoma and Shimura, for instance, examined the speech act of refusals by native speakers (NSs) of Japanese and American learners of Japanese, revealing several cases of negative transfer. The Japanese learners did not provide alternatives in refusals as frequently as the NSs did, and they were unable to use incomplete sentences that, in Japanese, serve to mitigate the face threat involved in refusals. Finally, previous research has focused on select features of discourse management skill as an index of pragmatic competence. One of the most studied features involves sentence-final particles

561

Naoko Taguchi (Ohta, 2001; Sawyer, 1992; Yoshimi, 1999, 2001). Sentence-final particles play an important role in Japanese communication because they help establish the interpersonal relationships among the interlocutors in interactions (Maynard, 1993). Ohta’s research documented the development of the use of the particle ne in acknowledgment expressions and expressions of alignment among beginning-level learners of Japanese. Analysis of naturalistic classroom discourse revealed that the teachers used a variety of listener responses including ne for acknowledgments, expressions of alignment, and confirmation questions and that the students gradually became able to internalize and adapt the particles in their interactions. Although these studies have provided a relatively rich representation of interlanguage pragmatics in Japanese, the literature has been largely limited to the areas of politeness, speech acts, and discourse features. Very little research has examined the comprehension of indirect meaning in L2/FL Japanese and thus it awaits future research. Features of Indirectness in Japanese Communication The bulk of previous research literature has unveiled features of indirectness in Japanese communication (e.g., Barnlund, 1989; Clancy, 1986; Davies & Ikeno, 2002; Gudykunst & Nishida, 1993; Imai, 1981; Klopf, 1991; Maynard, 1997, 1998; Miike, 2003; Miller 1994; Mizutani, 1981, 1985; Mizutani & Mizutani, 1987, 1988; Okazaki, 1993). This literature claims that the Japanese language, particularly its spoken form, makes indirect expressions possible through various syntactic and lexical features. One typical feature is the use of formulaic expressions to express dispreferred responses (e.g., refusals, disagreements). For example, in response to a request, one can say Kangaete okimasu ‘I’ll give it consideration.’ These kinds of expressions do not mean either yes or no but imply that there is little hope for the request (Imai; Miller; Mizutani, 1985). Some quantifier adverbs also convey negative implications: amari ‘not very,’ chotto ‘a little,’ and doomo (adverb of hesitancy1 ; Mizutani, 1985; Mizutani & Mizutani, 1987, 1988). Another example of indirectness in Japanese is the use of a questioning strategy when showing disagreement (Beebe & Takahashi, 1989; Furukawa, 2001; Kusakabe, 1994; Miller, 1994; Mizutani, 1985). Negative opinions are often marked with epistemic sentence-ending markers of hesitancy, such as: kana ‘I wonder,’ kamo shirenai ‘maybe,’ and to iu ki ga ‘I have a feeling.’ To ex-

press disagreement, making a question out of the prior speaker’s opinion is also common. For example, responding to the statement that Japanese education is getting better, one can challenge the opinion by saying Yoku naru? ‘Is it getting better?’ Other common strategies are partial agreement strategies, such as: Tashikani, demo ‘Surely, but’ and hesitation markers such as saa ‘I don’t know.’ Ellipsis is also a feature of indirect communication in Japanese (Fry, 2003). When giving a negative evaluation, the last part of the utterance is often left out to indicate the speaker’s reluctance to be negative (Miller, 1994; Mizutani & Mizutani, 1987, 1988). For example, when refusing an invitation, one could say, Sonohi wa ‘That day is . . . [not so good]’ A statement can be left incomplete when asking for a favor or making a suggestion. For instance, when inviting a guest to tea, a hostess can say Ocha ga hairi mashita kedo ‘The tea is ready, but,’ by omitting the subordinate clause following the conjunction kedo ‘but.’ These features of the Japanese language illustrate that implicit meaning is encoded by means of a variety of lexical and syntactic forms, as well as through formulaic expressions, which are used to convey reservations and disagreements indirectly. The present study adapted these conventional features of indirect communication to develop an instrument that measures pragmatic comprehension in JFL. In addition, the JFL learners’ comprehension of conventionalized implied meaning was compared with that of unmarked, nonconventional implicatures to interpret the findings in light of the previous findings in L2/FL English. The following two research questions guided the present study: 1. Does learner proficiency affect accuracy and speed in comprehending implied meaning in JFL? 2. Do JFL learners vary in accuracy and comprehension speed over different types of implied meaning? METHODOLOGY AND MEASURES Participants The participants consisted of 63 students enrolled in a Japanese language program in a U.S. university. Of these students, 30 were approaching the end of their second semester of Japanese study (elementary-level students or EJ students), and 33 students were nearing the end of their fourth semester of Japanese study (intermediate-level students or IJ students). The

562 EJ group had 14 males and 16 females, ranging in age from 18 to 39 years, with an average age of 21.43 years.2 None of the participants had experience living in Japan.3 In the EJ group, 17 were NSs of English, 7 were NSs of Chinese, 5 were NSs of Korean, and 1 was an NS of Spanish. The IJ group had 14 males and 19 females, ranging in age from 18 to 26 years, with an average age of 21.06 years. They averaged 2.4 years of formal Japanese study. Only one student had lived in Japan previously. Seventeen were English NSs, 9 were Chinese NSs, 6 were Korean NSs, and 1 was an NS of Thai. Both groups met for four class periods per week, 50 minutes each. On the background questionnaire, the EJ group reported studying Japanese 6.3 hours per week outside the classroom, whereas the IJ group reported about 6.9 hours per week of study outside the classroom. There was no significant difference in the amount of outside study time between the two groups (t = 0.71, p > .05). Japanese Pragmatic Listening Test (J–PLT) A computerized listening test called the Japanese Pragmatic Listening Test (J–PLT) was developed to measure the ability to comprehend implied meaning. The test had 50 items: 2 practice items, 12 filler items, and 36 experimental items. The filler items tested literal comprehension. The experimental items tested comprehension of three types of implied meaning: indirect refusals (12 items), conventional indirect opinions (12 items), and nonconventional indirect opinions (12 items). Each item consisted of a short conversation spoken by a male and a female NS of Japanese. In the experimental dialogues, the reply that appeared at the end did not provide a straightforward answer to the speaker’s question. Each conversation was followed by multiplechoice questions with four answer options (see the Appendix). The answer options were given in English because of the different levels of Japanese reading ability across the learner groups. All the option sentences were simple declarative sentences. The length of the conversations was kept relatively similar in order to control the impact on short-term memory. Each conversation had about 120 syllables, and from four to five speaker turns (about 22–25 seconds in length). The number of words in option sentences was kept approximately the same across items (28–30 English words). In order to reduce the extraneous effect coming from differences in linguistic knowledge among the learners, most vocabulary and

The Modern Language Journal 92 (2008) grammatical structures were drawn from the elementary Japanese course packet used at the target institution. The speech rate also was adjusted for the elementary level. Indirect Refusals. On the J–PLT, there were three types of indirect refusal items: refusing requests (4 items), refusing invitations (4 items), and refusing suggestions (4 items). The majority of the items were adapted from my (Taguchi, 2005) ESL version of the test. Utterances used for indirect refusals took the form of providing a reason and were written so as to avoid explicit negative linguistic markers such as “No, I can’t” or “I don’t want to.” Providing a reason or explanation is considered a common social practice in refusal situations (e.g., Nelson, Carson, Batal, & Bakary, 2002). Thus, indirect refusals reflect conventional language use, and a specific discourse pattern is customarily used in this type of refusal exchange. Conventional Indirect Opinions. The conventional indirect opinions on the test, by contrast, reflected the conventionality of language forms because the illocutionary meaning was conveyed through fixed lexical items or syntactic forms. When creating these items, the researcher consulted the literature on indirect communication in Japanese (see “Features of Indirectness in Japanese Communication” section). Three features of indirectness, which were cited by previous researchers, were adapted to this study: adverbs of reservation (three items), expressions of wondering (seven items), and reserved sentence endings (epistemic sentence ending markers used as hedging) (two items; see Table 1). Nonconventional Indirect Opinions. The expressions used in the nonconventional opinions on the test differed from the common refusal routines of providing an excuse or of using conventional linguistic markers to indicate negative opinions because they did not involve routine linguistic forms or language use patterns. An example of this type of expression is to indicate a negative opinion about the quality of an essay by saying “It is difficult to write an essay in Japanese, isn’t it?” As this example suggests, the linguistic options for expressing opinion are almost without limit, and they are more idiosyncratic and freer than for other implicature types. In order to counteract possible response biases, half of the items of this type were designed to illustrate negative opinions, and the other half were written to imply positive opinions (see the Appendix).

563

Naoko Taguchi TABLE 1 Linguistic Forms Used to Create Conventional Indirect Opinion Items Lexical and Syntactic Forms

Example

Degree Adverbs Indicating Reservation chotto ‘a little’ A: Frorida wa doo desuka? amari ‘not very’ (How do you like Florida?) doomo (hesitancy adverb) B: Frorida wa chotto. (Florida is a little . . . ) Making a Question out of Prior Speaker’s Opinion to Indicate Reservation A: Frorida wa ii desu ne. (Florida is nice.) B: Soo kana. (I wonder . . . ) Reserved Sentence Ending . . . kana to iu ki ga ‘I have a feeling about . . .’

A: Frorida wa doo desuka? (How do you like Florida?) B: Chotto atsu sugiru kana to iu ki ga. (I have a feeling that it’s too hot.)

Development of Distractor Options. The multiplechoice questions in the J–PLT had four answer options. When writing the distractor options, I applied three principles (Taguchi, 2005). These principles were used for a post hoc error data analysis to gain insights into potential problems when learners made an incorrect choice. The principles were as follows: Principle 1: The option contains a meaning opposite the implied meaning. Principle 2: The option contains words related to the last part of the conversation. Principle 3: The option is related to the overall conversation. The first principle was established to provide the least plausible competing source of information as a distractor. If learners have a sense of what is implied in the target utterance, this option would be the most remote choice for their response. The second principle was established based on the recency effect (i.e., the tendency for people to recall the words they heard last; Atkinson & Shiffrin, 1971). The third principle was established to see whether learners operate under the keyword processing strategy. Previous research has shown that weak listeners tend to go directly to the first association of the keyword and the referent, whereas proficient listeners tend to explore several possible associations between the words and referents before making a decision (Ross, 1997). Instrumentation and Pilot Study. When developing the J–PLT, I administered a written survey to

Representative Literature Imai (1981) Mizutani & Mizutani (1987, 1988)

Furukawa (2001) Imai (1981) Kusakabe (1994) Miller (1994) Mizutani & Mizutani (1988) Mizutani (1985) Mizutani & Mizutani (1987, 1988)

22 adult Japanese living in Japan. The survey asked them to report the instances of indirect communication that they had experienced or observed in their immediate lives. I also took field notes and recorded instances of indirect responses observed (e.g., indicating a negative opinion about someone’s presentation by saying “She [the speaker] seems very busy lately.”). A total of about 35 examples were collected, and the data were used to create nonconventional indirect opinion items. The draft items were piloted with two NSs who had backgrounds in Japanese linguistics, and both achieved 98% accuracy. Several items were revised based on their feedback. The final versions of the J–PLT items were computerized using the Revolution software for Macintosh and were piloted with 30 NSs of Japanese (16 males and 14 females). The purpose of the pilot test was to confirm comprehension accuracy because if NSs could not understand the meaning, the items would be considered inappropriate for JFL learners. Comprehension among the NSs was almost perfectly accurate for all item categories. The mean was 11.83 out of 12.00 for indirect refusals (SD = 0.46), 11.60 (SD = 0.67) for conventional indirect opinions, and 10.80 for nonconventional indirect opinions (SD = 0.53). Two problematic items were revised based on the results, and the accuracy of those items was confirmed with the same NS participants. As for comprehension speed, the hierarchical linear model revealed that the NSs’ comprehension response times differed significantly across item types (F = 6.42, p < .05) after controlling for their basic reading speed.4

564 Comprehension was fastest for indirect refusal items (M = 9.08, SD = 2.97). Although comprehension of conventional opinions (M = 10.15, SD = 2.66) was slightly faster than that of nonconventional opinions (M = 10.39, SD = 3.11), the difference was not statistically significant. After administering the J–PLT to JFL learners, the reliability of the instrument was assessed. The internal consistency reliability was estimated using Cronbach’s alpha, yielding .91 for the total implied meaning items (k = 36), .79 for indirect refusals (k = 12), .83 for conventional opinions (k = 12), and .74 for nonconventional opinions (k = 12).5 English Sentence Reading Task

The Modern Language Journal 92 (2008) introspective verbal interviews (i.e., by asking the learners to report their thoughts upon completing a task). After the J–PLT, 10 participants (5 EJ and 5 IJ students) were selected randomly and asked to participate in follow-up interviews in the researcher’s office. In the interviews, 12 items from the J–PLT (4 from each item type) were administered randomly, and the learners were asked to choose the correct answer. Then the verbal reporting session started. The researcher asked questions such as “Why did you choose the answer?” or “What were you thinking when you chose the answer?” The interview sessions, each lasting 20–30 minutes, were recorded with a digital voice recorder and later transcribed.

For this study, a separate English sentence reading task was also administered to the JFL learners for two reasons. First, the study examined the proficiency effect on the time taken to comprehend implied meaning correctly. Because the comprehension time included the time when the learners were reading the multiple-choice option sentences in English, it was necessary to control for the basic reading time in order to rule out the individual differences in reading speed. The second reason for a separate reading test was the mixed L1 backgrounds of the participant group. About 40–50% of the participants were NNSs of English. Although they were highly proficient in English (they had minimum scores of 600 on the Test of English as a Foreign Language [TOEFL]), there was no evidence that their reading speed was the same as that of NSs of English. Thus, a separate measure was necessary to control for baseline reading speed. The sentence reading task had a total of 25 simple English sentences of either seven or eight words (e.g., “The man is taking a business class.”). The structures of the sentences mirrored those of the multiple-choice options, but vocabulary items were replaced as necessary to avoid a repetition effect. The participants read each sentence, and their reading time was recorded. To ensure comprehension, yes/no questions (for a total of five) asking about the content of a sentence were inserted periodically. The accuracy rate of the five comprehension questions was 96.5%, indicating that the participants were comprehending the sentences for meaning. None of them made more than one mistake.

DATA COLLECTION PROCEDURES

Introspective Verbal Interviews

The first research question focused on the effect of general proficiency on the accuracy and speed of pragmatic comprehension. Proficiency was operationalized as levels of study (EJ and

The learners’ internal processes during pragmatic comprehension were investigated by

The J–PLT was administered using Macintosh computers in a computer lab on campus. During the test, the learners wore headphones and read instructions in English on the computer screen. Before starting the test items, they practiced 2 items. Immediately following each conversation, a written multiple-choice question appeared on the screen. The participants chose the most appropriate response by pressing the corresponding number on the keyboard. Once they chose an answer, the computer automatically took them to the next item. Response time was measured between the moment that the question appeared on the screen and the moment that the learners pressed the number key. Immediately following the J–PLT, the learners completed the English sentence reading task using the same computer. They read instructions on the screen and practiced 2 items. Then they proceeded to the 25 target sentences. They were told to read each sentence on the computer screen at a normal speed and press the 1 key as soon as they finished reading. They were also asked to respond to the yes/no comprehension question that appeared occasionally. Approximately 1 week after administration of the J–PLT, 10 learners (5 EJ and 5 IJ) had introspective verbal interview sessions in the researcher’s office. The recorded verbal protocols were transcribed and analyzed qualitatively for notable patterns in the learners comprehension processes. DATA ANALYSIS

565

Naoko Taguchi IJ). Comprehension accuracy was measured by the J–PLT, which had an interval scale between 0 and 36 (12 items per item type). Comprehension speed was operationalized as response times and was calculated by averaging the number of seconds taken to answer items correctly.6 A multivariate analysis of variance (MANOVA) assessed the proficiency effect on accuracy scores of three item types, and a multivariate analysis of covariance (MANCOVA) was used to examine the proficiency effect on response times, with English sentence reading time as a covariate. Because the distribution of response times did not confirm the assumptions of normality, a logarithmic transformation was performed (Tabachnick & Fidell, 2001).7 After the transformation, the distributions were normal and linear, and homoscedaticity was confirmed. The alpha level was set at .05. The second research question asked whether accuracy and comprehension speed differed across the types of implied meaning. Differences in scores and response times among the three item types were examined by using a hierarchical linear model (HLM), with the basic reading speed measured by the sentence reading task as a covariate. An HLM was used in this study instead of a general linear model because it works for repeated measures data (data taken from the same individuals over multiple times) and allows a more direct interpretation of the role of covariate. An HLM also was considered appropriate for this study because the data were collected from intact classes without random sampling and the observations were not fully independent of each other (Osborne, 2000; Raudenbush & Bryk, 2002). The alpha level was set at .05.

RESULTS Effect of L2/FL Proficiency on Comprehension of Implied Meaning The first research question asked whether proficiency affected accuracy or comprehension speed in response to different types of implied meaning. Tables 2 and 3 display descriptive statistics of the scores and response times, respectively. The MANOVA revealed that the IJ group was significantly more accurate in comprehending all types of implied meaning than the EJ group: F = 18.62 for indirect refusals, F = 14.23 for conventional indirect opinions, and F = 30.64 for nonconventional indirect opinions (p < .05). The effect size was moderate: η2 of .23, .19, and .33, respectively.8 The proficiency effect on comprehension time was examined by applying the MANCOVA. After adjusting for the covariate (i.e., English reading time), the EJ and IJ groups did not differ in response times, F = 1.26 (ns). Hence, there was a strong effect of proficiency on the accurate comprehension of implied meanings of all types, whereas proficiency did not influence the comprehension speed of any of the implied meaning item types.9 Effect of Implied Meaning Types on Comprehension The second research question asked whether L2/FL learners differ in the accuracy and comprehension speed shown for the three different types of implied meaning. As shown in Table 2, there was a variation in accuracy scores across item types. For both the EJ and IJ groups, indirect refusal was the easiest to comprehend, and conventional

TABLE 2 Descriptive Statistics for Comprehension Scores Group and Item Types Elementary Japanese Group (n = 30) Total Indirect Refusals Conventional Indirect Opinions Nonconventional Indirect Opinions Intermediate Japanese Group (n = 33) Total Indirect Refusals Conventional Indirect Opinions Nonconventional Indirect Opinions

K

M

SD

Min.

Max.

36 12 12 12

20.57 8.50 5.90 6.17

6.02 2.46 2.25 2.26

9.00 4.00 2.00 1.00

32.00 12.00 11.00 10.00

36 12 12 12

27.79 10.70 8.03 9.06

4.72 1.51 2.23 1.87

14.00 5.00 2.00 5.00

34.00 12.00 11.00 12.00

Note. Total refers to all items from three item categories combined. K refers to the number of items in each category. One point was given per correct answer.

566

The Modern Language Journal 92 (2008)

TABLE 3 Descriptive Statistics for Comprehension Speed (Response Times) Group and Item Type Elementary Japanese Group (n = 30) Total Indirect Refusals Conventional Indirect Opinions Nonconventional Indirect Opinions Intermediate Japanese Group (n = 33) Total Indirect Refusals Conventional Indirect Opinions Nonconventional Indirect Opinions

K

M

SD

Min.

Max.

36 12 12 12

12.16 10.41 12.90 14.25

3.63 2.93 6.50 5.15

5.48 4.22 5.68 5.87

21.10 16.80 39.90 27.54

36 12 12 12

11.50 9.73 12.22 14.72

3.32 2.79 5.98 4.53

5.90 5.08 4.98 6.06

19.88 15.95 33.37 26.72

Note. Total refers to all items from three item categories combined. Response time in seconds refers to the average number of seconds taken to comprehend each item correctly. K refers to the number of items in each category.

indirect opinions were the most difficult. The HLM analysis confirmed that the difference was statistically significant across item types at the alpha level of .05: F = 29.52, p < .05 for the EJ group and F = 35.15, p < .05 for the IJ group. The effect size was also large: η2 of .67 for the EJ group and .64 for the IJ group. For the IJ students, accuracy scores for all item type pairs differed significantly. They were most accurate in comprehending indirect refusals, followed by nonconventional indirect opinions. For them, comprehension of conventional indirect opinions was the most difficult. These patterns were the same for the EJ group except for the comparison between conventional and nonconventional indirect opinions, which did not yield a significant difference. As shown in Table 3, for both JFL groups, response time was the shortest for comprehending indirect refusals, followed by the comprehension of conventional indirect opinions. Both groups needed the longest time to comprehend nonconventional indirect opinions. The HLM analysis was applied to see if the difference observed in pragmatic processing was statistically significant. Basic processing speed, obtained from the English sentence reading task, was used as a covariate. The difference was found statistically significant: F = 13.24, p < .05 for the EJ group, and F = 28.80, p < .05 for the IJ group.10

taken from the last heard utterance of the dialogue or their synonyms, and (c) an option containing meaning related to the overall conversation. As shown in Table 4, the IJ learners were most attracted to the second distractor type, and the recency effect became more discernible as item difficulty increased. When comprehending conventional indirect opinions (the most difficult item type), the IJ learners chose the distractor that contained the words taken from the last utterance more often that the other distractors. Hence, it seems that when implied meaning was less accessible to the IJ students, they tended most often to rely on their memory when making choices. The EJ students, in contrast, were attracted to both the first and the second distractor options in a similar manner, which reflected their limited proficiency. The first option provided the least plausible competing source of information as a distractor because it contained a meaning opposite the target implied meaning. The proportion of the EJ students who were attracted to this distractor was greater than that of the IJ students, indicating that the EJ students’ listening proficiency was not strong enough to recognize the remoteness of this distractor type.

Analyses of the Introspective Verbal Reports Post Hoc Analyses of Error Data in Comprehension The error data were examined for the response options of each multiple-choice question in the J–PLT. The three distractor options were (a) an option containing a meaning opposite the implied meaning, (b) an option containing words

The present study used follow-up introspective verbal interviews to examine the learners’ internal processes when comprehending implied meaning. For the study, 120 verbal reports (12 items per learner; 10 learners) were analyzed for incidences of particular strategies used to derive inferences from implied meanings. When

567

Naoko Taguchi TABLE 4 J–PLT Distractor Analysis Item Type and Group Indirect Refusals Elementary Japanese Group Intermediate Japanese Group Conventional Indirect Opinions Elementary Japanese Group Intermediate Japanese Group Nonconventional Indirect Opinions Elementary Japanese Group Intermediate Japanese Group Total Average Elementary Japanese Group Intermediate Japanese Group

Percent Correct Responses

Distractora

Distractorb

Distractorc

71.2% 89.5%

16.7% 7.2%

8.8% 2.8%

3.3% 0.6%

45.0% 60.8%

20.6% 5.8%

27.8% 24.0%

7.5% 2.8%

51.4% 76.1%

17.7% 6.6%

17.3% 10.7%

17.3% 6.6%

54.8% 77.2%

18.3% 6.7%

17.9% 12.7%

9.3% 3.4%

Note. There were 30 students in the Elementary Japanese group and 33 students in the Intermediate Japanese group. Numbers in the table show percentages for each option chosen by the group. a Distractor meaning opposite to the implied meaning. b Distractor meaning related to last heard utterance. c Distractor meaning related to overall conversation.

comprehension was relatively easy, the learners tended to use very few inferential strategies during comprehension. This was typically the case for indirect refusals. For example, in Item 37, the woman speaker refused the man’s invitation to go to the movie by saying “I have a test tomorrow.” Out of 10 learners, 9 responded that the woman could not go to the movie because she had to study for the test. This finding shows that the learners did not have to use many strategies to understand the illocutionary intent of the refusals, probably because they were embedded in conventional, easily recognizable contexts. When the learners did not understand the target utterance, they used various cues to draw inferences. A common paralinguistic cue was the speaker’s tone and intonation. Out of the 120 reports, there were 21 instances for which the learners noticed a sign of hesitation in the speaker’s voice and related it to negative meaning. For instance, Excerpt 1 from an EJ learner’s interview shows that although he did not understand the meaning of the adverb doomo as an adverb of hesitancy, but he was able to make the correct inference by using intonation as a relevant cue: EXCERPT 1 Item 23 (The woman asks the man if he likes the teacher; the man responds, Ano sensei wa doomo ‘That teacher is [hesitancy adverb].’) Student: Ano sensei wa doomo, he said it in a kind of long, extended tone. I don’t know what

doomo means, but it’s kind of sarcastic? Doomo usually means ‘very,’ but the way he said it was negative, ‘very bad,’ but he didn’t want to say ‘bad.’ Another notable strategy that the learners used was to rely on their knowledge of adjacency pair patterns. Conversation has a sequence of turns, and normative patterns of conversation serve as a frame of reference when understanding each participant’s response (Goffman, 1976; Sacks, Schegloff, & Jefferson, 1974). This concept explains how the first pair-part determines what is relevant in the next turn (i.e., second pair-part). For example, a greeting sets up the expectation that the next pair-part should be a return greeting. A question asking for evaluative comments should be followed by “good” or “bad.” According to Levinson (1983), the second pair-part often falls into a preferred or dispreferred response category. For instance, saying “sure” when accepting an invitation is a preferred response, whereas refusing it with an explicit “no” is a dispreferred reply. Preferred second turns are unmarked and occur as structurally simple turns, whereas dispreferred seconds are complex and marked with various linguistic and nonlinguistic devices to mitigate the potential face threat (e.g., pauses, fillers, token agreements, affective devices, and reasons why the preferred second cannot be performed). Because knowledge of the adjacency pair pattern and dispreferred seconds is shared between Japanese and English, it facilitates the learners’ comprehension of implicit meaning in the target language. As shown in

568 Excerpt 2, the learner took advantage of this shared discourse convention to seek the meaning of the indirect reply: EXCERPT 2 Item 12 (The man asks the woman if she likes the pasta, and she responds Kore wa cheezu ga totemo ooi desu ne ‘It’s got a lot of cheese in it.’) Student: Smith-san said this restaurant is famous, so she is obligated to say something nice, but if the only thing she can come up with is the cheese, she must not like it so much. It’s the same in English too. When you get a sweater you don’t like and if you say, “This sweater is very red,” it’s not a compliment. So the comment she said was too superficial. It’s not Oishiiwayo ‘It’s delicious,’ or something interesting about the pasta. Finally, knowledge of linguistic conventions was also evident in both L2 groups, particularly the knowledge of the adverb chotto ‘a little,’ which conveys a negative tone. Out of 10 learners, 4 elaborated on the use of chotto in Japanese communication and how it marks negative connotation: EXCERPT 3 Item 24 (The man asks the woman if she wants to stay for dinner, and she responds Chotto osoinde ‘It’s a bit late so . . .’) Student: She used chotto for disagreement. Chotto is polite refusal. I learned it from my teachers. When I say something not right, they say Ma chotto in class. It’s not negative, but it’s polite negative. So it’s not like I really disagree with you, but it’s more like “well” to me, chotto, is “well,” in English. So it’s the same like, “Do you like this picture?” and you say “Well.” It’s a polite way to say it’s not so good, but saying it nicely. Chotto is supposed to mean “a little,” so I picked that up when my teacher says Chotto . . . kamoshirenaine, meaning “I don’t know about that,” so I picked that up from her. Here, the IJ learner was able to distinguish the literal meaning of chotto ‘a little’ from its conventional, negative function and used the knowledge to draw inferences. Some linguistic conventions were salient, as in the case of chotto, but others were left unnoticed. For example, in Item 29, the male speaker expressed a negative opinion about

The Modern Language Journal 92 (2008) a picture by saying Kuruma ga chotto ookii kana to iu ki ga ‘I have a feeling that the car might be a little bit big.’ No learners relied on the reserved ending expression (to iu ki ga) to draw a negative inference, but they focused exclusively on the content meaning (‘The car is big.’). Findings from the present study revealed different degrees of saliency among conventional linguistic forms, probably due to the frequency of their occurrence in daily interactions, as seen in the learners’ verbal protocol data. More IJ students than EJ students demonstrated explicit knowledge of those pragmalinguistic cues. They were able to pick up specific phrases and grammatical markings used to convey indirect meaning and articulate their pragmatic functions, whereas such instances were rare in the EJ students’ reports. One example is the case of Doo deshoo ne ‘I wonder,’ an expression of wondering used to show disagreement. Whereas all of the EJ learners interpreted it literally, as a sincere question, the majority of the IJ students were able to point out the function of this rhetorical question as a polite disagreement: EXCERPT 4 Item 13 (The man asks the woman if it is a good idea to major in Japanese, and she replies Mejaa wa doo deshoo ne, meaning “I wonder about the major.”) Student: Doo deshoo. I remember we talked about it in the language and culture class. It shows reservation of opinion. In English we would say “I don’t know about it” if it’s negative. She doesn’t think a Japanese major is a good idea. A similar example is the questioning strategy that indicates disagreement—Soo desu ka? ‘Is that so?’ Out of five IJ students, three students were able to distinguish the two uses of this expression: Soo desu ka with falling intonation showing acknowledgment and Soo desu ka with rising intonation showing disagreement. They noticed the rising intonation of this expression and successfully comprehended the negative intention attached to it. No EJ students verbalized these rules. EXCERPT 5 Item 8 (The man says that Boston has better museums than New York, and the woman responds Soo desu ka?) Student: She did not say she likes Boston museums better. . . . Soo desu ka with falling

Naoko Taguchi intonation is ‘I see.’ With rising intonation it’s a question. Hers is a question, but I guess it’s a sort of disagreement. In summary, the introspective verbal interviews revealed a range of strategies that the learners used during the task of inferencing (e.g., paralinguistic cues, knowledge of adjacency pairs, knowledge of pragmalinguistic conventions). Qualitative analyses revealed (a) the types of contextual and cognitive cues that were most immediately accessible to the learners when they were seeking the relevance of implicit input and (b) proficiency effects on the knowledge of pragmalinguistic cues. DISCUSSION The Effect of Proficiency on Pragmatic Comprehension in Japanese This study revealed a significant effect of proficiency on the accuracy of comprehension for all three types of implied meaning. Regardless of the degree of conventionality, or of the difficulty levels across item types, learners with higher proficiency were always more accurate in comprehending implied meaning than lower proficiency learners were. These findings, then, replicated the previous findings in L2 English (Cook & Liddicoat, 2002; R¨over, 2005; Taguchi, 2002, 2005, 2007), lending support to crosslinguistic generalizability: A proficiency advantage exists in the comprehension of indirect meaning in L2/FL Japanese. As shown in the post hoc error analysis, when the learners made an incorrect choice, they were most distracted by the distractor option that contained words from the last utterance heard, which is suggestive of the recency effect in listening. Because the input in listening is acoustic and transient in nature, listening requires automatic, realtime comprehension. Listening is greatly affected by short-term memory because after the information is heard, there is no text to refer back to. In the present study, it seems that when a target meaning was not easily accessible, the learners were distracted by the information that they remembered most easily. For the less proficient learners, however, the distracting cues were not solely in the last utterance heard. Their limited proficiency was evident in that they also had a tendency to choose the most remote distractor (the meaning opposite the target meaning) more often than the higher proficiency group. Although accuracy positively correlated with proficiency, proficiency had no effect when the speed of comprehension was analyzed. Higher proficiency learners took about the same amount

569 of time as lower proficiency learners to process pragmatic information, to read answer options, and to choose the correct interpretation, and this tendency was the same across item types. Answer options were given in English, the language in which the learners were proficient, and the researcher controlled individual differences in reading speed by treating it as a covariate. Hence, response times closely reflected the level of the learners’ Japanese proficiency: the ability to understand conversations, remember the content, and draw inferences. However, the Japanese proficiency was not a factor in speedy comprehension of implied meaning in this study. These findings support my previous findings on pragmatic comprehension in EFL (Taguchi, 2005, 2007). In these prior studies, EFL learners at different proficiency levels took a listening test that measured their ability to comprehend conversational implicatures. L2/FL proficiency, operationalized as the institutional TOEFL scores, had a significant influence on accuracy (scores) but not on comprehension speed (response times). I (Taguchi, 2005) provided several explanations for the findings. The first explanation pointed out the difference between lower and higher order processing. Previously, response time measures were used mostly to examine lower order, componentlevel processes (e.g., word recognition, sentence reading) to make inferences about fluent control of language processing (e.g., Jiang, 2002; Segalowitz, 2000; Segalowitz, Segalowitz, & Wood, 1998). These studies confirmed that speedy processing of linguistic information results from the automatization of underlying component processes, which is related to general proficiency. Higher order processing, such as pragmatic processing, by contrast, goes beyond lower order processing because it involves processing at multiple levels. Lower order processing (e.g., perception of sounds and recognition of words) is prerequisite, but at the same time, learners need to use pragmatic knowledge that encompasses a wide range of properties, including knowledge of relevance, conventions of language use, and sociocultural norms of interaction. Learners need to attend to a range of contextual cues simultaneously, including paralinguistic cues (e.g., intonation and pause), discourse cues (e.g., patterns of adjacency pairs), and cognitive cues (e.g., background knowledge). Because pragmatic processing involves a number of component processes that need to be automatized to achieve speedy processing, a proficiency effect may not be evident among the beginning- and intermediatelevel learners studied here.

570 A second potential explanation is the context of learning. Similar to the participants in my earlier studies (Taguchi, 2005, 2007), the learners in the present study were foreign language learners who had limited contact with the target language outside the classroom. Improved speed in pragmatic processing simply may be due to repeated processing practice and exposure to input, which neither JFL group in this study had extensively in the course of their Japanese studies. The participants in this study were enrolled in a formoriented Japanese class that met only four times a week, for 50 minutes each time. In the background survey, they reported spending about 6– 7 hours per week studying Japanese outside the classroom, mostly working on grammar, vocabulary, and kanji. This amount of time was similar between the IJ and EJ groups. Hence, it may be the function of the amount of processing practice and time on task, rather than the proficiency per se, that determines the increased speed in pragmatic comprehension. These interpretations correspond to the theories of skill development (e.g., Anderson & Lebiere, 1998; McLaughlin, 1987). Speedy skill execution is promoted through consistent associative practices between input and responses, and performance speed develops as these associations become strengthened by repeated activation. Exposure to input and abundant processing practice are necessary conditions for speed development. These theoretical claims are, in part, supported by the post hoc correlational analyses between comprehension speed and the amount of language contact hours reported by the learners on the background questionnaire. The Pearson’s correlation revealed a significant relationship between response times and the amount of time that the participants devoted to Japanese outside the classroom: r = −.37 (p < .01) for comprehension speed of indirect refusals, r = −.32, (p < .05) for comprehension speed of conventional indirect opinions, and r = −.33 (p < .01) for comprehension speed of nonconventional indirect opinions. In contrast, the accuracy scores bore no relationship to the amount of Japanese contact reported. These findings support the claim that time on task affects the speedy processing of pragmatic information. Similar findings were reported in my (Taguchi, 2008b) study, which compared the development of accurate and speedy comprehension of implied meaning between ESL and EFL learners. The ESL learners showed more profound development in the speed of pragmatic comprehension than the EFL learners. Another

The Modern Language Journal 92 (2008) study (Taguchi, 2008a) also showed that ESL learners’ gains in comprehension speed correlated significantly with the amount of language contact outside the classroom. Again, the present study, conducted in the context of JFL, lends support to those previous findings.

Effect of Item Type on Pragmatic Comprehension The second research question addressed the effect of implied meaning types on comprehension accuracy and speed. For the IJ group, there was statistical evidence that indirect refusals were the easiest and fastest to comprehend. Comprehension of conventional indirect opinions was less accurate than that of nonconventional indirect opinions; however, comprehension speed was faster for the conventional opinions than for the nonconventional opinions. These findings suggest that although some linguistic conventions encoded in the items were difficult for the IJ learners to access, when they understood the conventions they were able to derive meanings quickly by relying on the conventions. Nonconventional items, by contrast, did not reveal such a facilitative effect in comprehension speed. The EJ group demonstrated the same patterns as the IJ group, except that the conventional and nonconventional indirect opinions were equally difficult for them (both item types produced low accuracy scores), probably because of their lower proficiency. The findings of the present study for JFL largely replicated the previous findings for L2/FL English (Carrell, 1984; Cook & Liddicoat, 2002; R¨over, 2005; Taguchi, 2002, 2005, 2007, 2008a). Different comprehension demands exist across implied meaning types in L2/FL Japanese, reflecting the different degrees of conventionality involved, and the differences manifest in both accuracy and speed of comprehension. In the present study, however, the observed effect of item type on comprehension speed was particularly notable because it controlled for the individual differences stemming from general processing speed by using the basic sentence reading time as a covariate. Hence, the comprehension speed analyzed here more closely reflected the pragmatic portion of the processing speed than the general language processing speed. The faster response times for indirect refusals and conventional opinions suggest that the features of conventionality involved in the pragmatic processing helped to reduce the processing effort and speed up overall comprehension.

571

Naoko Taguchi The indirect refusals and conventional indirect opinions used in this study parallel two types of conventionality proposed by Morgan (1978): conventions of usage and conventions of language. Indirect refusals reflect conventions of usage because they include a common pattern of discourse in the refusal routine (i.e., giving an excuse). Because the pattern is frequent and customary, its meaning is easily accessible to L2/FL learners, particularly when the convention is shared between their dominant and target languages, as in the case of English and Japanese studied here. The indirect refusals were the easiest for both L2/FL groups to comprehend because the learners were able to apply familiar communication patterns from their L1s to the L2/FL. The introspective verbal reports also supported this tendency. Neither learner group used many strategies to comprehend indirect refusals; instead, they were able to articulate the motive behind the refusals (i.e., the speaker gave an excuse to avoid being too direct and impolite). As evidenced in these instances, illocutionary force was transparent and easily recognized in the refusals because they were already embedded in the routine pattern of social interaction. The conventional indirect opinions, however, reflected conventions of language. Conventionality was encoded through specific lexical and linguistic forms, such as adverbs of negative implications (e.g., chotto ‘a little’) and reserved sentence endings (e.g., N/Adj/V + kana to iu ki ga ‘I have a feeling about’). These conventional features are believed to assist comprehension because they directly mark meaning and are understood as almost formulaic (Sperber & Wilson, 1995; Verschueren, 1999). When speaker intentions are linguistically coded or embedded within predictable patterns of discourse, the listener may not attend to many contextual cues, which consequently reduces the processing effort. Similarly, in L2/FL comprehension, when learners are familiar with conventional forms, meaning becomes transparent if they are skilled enough to comprehend the utterances. For both groups, comprehending the conventional opinions was more difficult than understanding refusals, probably because certain conventional forms used in this study were unfamiliar to them. The introspective verbal interviews revealed that some quantifying adverbs (e.g., doomo, adverb of hesitancy) were unfamiliar to the learners. The questioning strategy for disagreement (e.g., saying “Is that so?”) was also sometimes mistaken for a sincere question. Although the learners relied on other strategies (e.g., into-

nation, hesitation) to compensate for their lack of knowledge, their unfamiliarity with the forms seems to have been the source of difficulty in comprehension. In the present study, nonconventional opinions were more difficult for the learners to comprehend than refusals, probably due to their idiosyncratic nature. Meaning in these implicatures was not attached to specific linguistic expressions, nor was it stable across instances of language use or users. As a result, the processing effort for these items was greater than for other item types because the learners needed to process a large number of contextual cues in order to understand meaning. Due to the broader contextual cues to be processed, along with the open set of possible expressions, comprehending the nonconventional opinions required greater processing effort from the L2 learners than for understanding refusals. This tendency manifested more strongly for the EJ group than for the IJ group in this study because the accuracy scores for the EJ group were equally low for both the conventional and the nonconventional indirect opinions, compared with indirect refusals. As shown in the verbal protocols, when comprehending the nonconventional items, the learners relied on a variety of contextual cues that were salient to them, including paralinguistic cues (e.g., intonation, hesitation) and knowledge of adjacency pair patterns. Different strategies identified in the verbal protocols lend support to the relevance theory (Sperber & Wilson, 1995) that utterance interpretation is neither a mere decoding of linguistic input nor a retrieval of logic. Various kinds of information in context, both linguistic and nonlinguistic, are placed on the same level to assist comprehension. These contextual cues assist learners’ comprehension when the target utterances lack conventional forms or the conventions are unfamiliar to the L2/FL learners. LIMITATIONS OF THE STUDY AND IMPLICATIONS FOR FUTURE RESEARCH The present study developed an original instrument to measure pragmatic comprehension in JFL and partially confirmed the previous findings for L2/FL English. Future research is needed over different participant populations to confirm the generalizabiliy of the findings. Inclusion of advanced-level learners is particularly important because this study did not find a proficiency effect on the speed dimension of pragmatic comprehension. Although the present study used the length of Japanese study as an indicator of proficiency,

572 an independent proficiency measure in the form of standardized exams or teachers’ ratings would be useful to assess the proficiency effect on comprehension precisely. Future research is needed over different target languages to expand the generalizability of the findings. The instrument developed in this study could be applied to address a variety of still underexplored issues in pragmatic comprehension. One important issue relates to the development of pragmatic comprehension. Previous longitudinal studies in L2 English have identified a hierarchy of difficulty and differential patterns of development between accuracy and comprehension speed. Learners find it easier to comprehend some items with implied meanings than other items, and accuracy gains precede speed gains. It would be interesting to examine whether these patterns of development found in L2 English are also present in other L2s. Such research will help extend the crosslinguistic generalizability of the findings and, at the same time, contribute to the rather neglected area of developmental research in pragmatic comprehension. Another limitation of the present study relates to the construct of pragmatic comprehension and to the methods used to develop the instrument. This study developed test items largely based on a literature review, survey, and field notes. Although the linguistic features gleaned from these methods were useful in writing indirect utterances, additional measures could enhance the authenticity and validity of the items. For instance, the survey could be administered to a larger number of Japanese NSs in order to compile more baseline situations. After compiling a pool of situations, the degree of conventionality involved in different situations could be assessed by having NSs rate each situation for the perceived degree of familiarity. Another potentially useful method is to analyze a corpus of naturally occurring conversations. Adjacency pairs in conversations can be analyzed for instances of dispreferred responses in order to identify the linguistic means used to convey disagreement or refusal intentions. If these methods are applied across languages, the process of item development will also help us understand the typological organization of conversational implicatures. It will help us to understand what types of implicatures exist in different languages and to identify the similarities and differences in the ways of communicating those implicatures crossculturally. Finally, the present study found a proficiency effect on pragmatic comprehension but did not reveal other factors that affected comprehension,

The Modern Language Journal 92 (2008) particularly factors related to instruction. Because the learners were all FL learners, their classroom experience probably shaped much of their learning experience. An interesting post hoc analysis would be to look at the instructional materials to see what types of inferential processing were practiced in class. We could also look into the textbooks for the three types of implicatures included in the instrument to see whether the different comprehension accuracy found across item types can be explained from the textbook materials. Numerous instructional studies have revealed that certain pragmatic features are teachable and that explicit instruction could facilitate the learning of those features (Rose & Kasper, 2002; Soler & Flor, 2005). Thus, classroom materials will form a valuable source of information in understanding the factors that affect the development of pragmatic comprehension.

ACKNOWLEDGMENTS I am grateful to Sono Takano and Yasufumi Iwasaki for their feedback and assistance with the listening instrument. I am also deeply indebted to Marc Siskin for his technical support on the instrumentation and data collection. Thanks also go to Howard Seltman for his advice on the statistical analyses. I am solely responsible for all the errors that may remain.

NOTES 1 Doomo is an adverb of hesitancy. It is used when one cannot make a definite statement about something due to lack of hard evidence (Makino & Tsutsui, 1997). 2 There was only one student over age 30; the rest were between 18 and 22 years old. 3 One student in the EJ group had 1 year of prior Japanese study before entering the university, but she was placed into the EJ group because of her scores on the placement exam. 4 The multiple-choice options were given in Japanese to the Japanese NSs. Their basic reading speed was measured via a separate sentence reading task that had 25 Japanese sentences. The participants read each sentence and pressed the number 1 key to move on to the next sentence. Their average reading time was entered as a covariate to the hierarchical general model. 5 k refers to the number of items in each category. 6 Response times involved the time spent to make pragmatic inferences based on the conversation just heard, to read sentence options in English, and to choose the right option that best reflected the conversation. The time spent on reading English sentences was taken into consideration by measuring the basic reading speed and using it as a covariate.

Naoko Taguchi 7 Transformation

is one way to fit data more closely to the underlying assumptions of statistical tests. Because the data were positively skewed, a logarithmic transformation was performed. The procedures followed Taguchi (2005), in which I consulted with professional statisticians. Similar procedures were used in previous studies (Juffs & Harrington, 1996). 8 There was a significant group difference on the comprehension of literal items with large effect size (t = –3.25, p = .002, η2 = .80). Because the literal items measured the learners’ global comprehension ability, not their inferential skill, the significant performance difference suggests that the two groups differed in their general listening proficiency, which lends support to the proficiency difference between the two groups. 9 Comprehension accuracy and response times were compared between Asians (Chinese, Koreans, and Thai) and non-Asians (Americans and Spanish). The Asian students were significantly more accurate than non-Asians: t = –2.94, p < .05 for the EJ group, and t = –2.13, p < .05 for the IJ group. There was no significant group difference in response times. However, this result should be interpreted with caution because the Asian participant group included heritage speakers and students who had spent a number of years in the United States. One participant in the IJ group had spent one semester in Japan previously. His accuracy score on the listening test ranked 7th out of 63, and his comprehension time ranked 13th. 10 The difference in response times across item types was significant without controlling for the covariate (reading time), F = 13.24, p = .00 for the EJ group and F = 35.15, p = .00 for the IJ group.

REFERENCES Anderson, J. R., & Lebiere, C. (1998). The atomic components of thought. Mahwah, NJ: Erlbaum. Atkinson, R. C., & Shiffrin, R. M. (1971). The control of short-term memory. Scientific American, 225, 82– 90. Barnlund, D. (1989). Communicative styles of Japanese and Americans. Belmont, CA: Wadsworth. Beebe, L. M., & Takahashi, T. (1989). Sociolinguistic variation in face-threatening speech acts: Chastisement and disagreement. In M. R. Eisenstein (Ed.), The dynamic interlanguage: Empirical studies in second language variation (pp. 199–218). New York: Plenum. Bezuidenhout, A., & Cutting, C. (2002). Literal meaning, minimal propositions, and pragmatic processing. Journal of Pragmatics, 34, 433–456. Bouton, L. (1992). The interpretation of implicature in English by NNS: Does it come automatically without being explicitly taught? Pragmatics and Language Learning Monograph Series, 3, 64–77. Bouton, L. (1994). Can NNS skill in interpreting implicature in American English be improved through

573 explicit instruction? A pilot study. Pragmatics and Language Learning Monograph Series, 5, 88–108. Bouton, L. (1999). Developing nonnative speaker skills in interpreting conversational implicatures in English. In E. Hinkel (Ed.), Culture in second language teaching and learning (pp. 47–70). Cambridge: Cambridge University Press. Carrell, P. L. (1984). Inferencing in ESL: Presuppositions and implications of factive and implicative predicates. Language Learning, 34, 1–19. Clancy, P. M. (1986). The acquisition of communicative style in Japanese. In B. Schieffelin & E. Ochs (Eds.), Language socialization across cultures (pp. 213–250). Cambridge: Cambridge University Press. Cook, H. (2001). Why can’t learners of JFL distinguish polite from impolite speech styles? In G. Kasper & S. Ross (Eds.), Pragmatics in language teaching (pp. 80–103). Cambridge: Cambridge University Press. Cook, M., & Liddicoat, A. J. (2002). The development of comprehension in interlanguage pragmatics: The case of request strategies in English. Australian Review of Applied Linguistics, 25, 19–39. Davies, R., & Ikeno, O. (2002). The Japanese mind. Tokyo: Tuttle. Fry, J. (2003). Ellipsis and wa-marking in Japanese. New York: Taylor & Francis. Furukawa, K. (2001). American JFL learners’ strategies in stating opinions. Unpublished master’s thesis, University of Oregon, Eugene. Garcia, P. (2004). Developmental differences in speech act recognition: A pragmatic awareness study. Language Awareness, 13, 96–115. Gibbs, R. W. (1981). Your wish is my command: Convention and context in interpreting indirect requests. Journal of Verbal Learning and Verbal Behavior , 20, 431–444. Gibbs, R. W. (1983). Do people always process the literal meanings of indirect requests? Journal of Experimental Psychology: Learning, Memory, and Cognition, 9 , 524–533. Goffman, E. (1976). Replies and responses. Language in Society, 5, 254–313. Gudykunst, W. B., & Nishida, Y. Y. (1993). Interpersonal and intergroup communication in Japan and the United States. In W. Gudykunst (Ed.), Communication in Japan and the United States (pp. 149–214). Albany: State University of New York Press. Hamblin, J., & Gibbs, R. (2003). Processing the meaning of what speakers say and implicate. Discourse Processes, 35, 59–80. Holtgraves, T. (2007). Second language learners and speech act comprehension. Language Learning , 57 , 596–610. Ikoma, T., & Shimura, A. (1993). Pragmatic transfer from English to Japanese: The speech act of refusals. Journal of Japanese Language Teaching , 79 , 41–52. Imai, M. (1981). Sixteen ways to avoid saying no. Tokyo: Nihon Keizai Shinbun.

574 Jiang, N. (2002). Form–meaning mapping in vocabulary acquisition in a second language. Studies in Second Language Acquisition, 24, 617–637. Jones, R. (1998). A comparative study of strategies used by native and non-native speakers of Japanese in responding to compliments. Studies in Japanese Language Teaching , 5, 54–71. Juffs, A., & Harrington, M. (1996). Garden path sentences and error data in second language sentence processing. Language Learning , 46 , 283–326. Kashiwazaki, H. (1993). Hanashikake koudou no danwa bunseki: Irai youkyuu hyougen no jissai o chuushin ni [Discourse analysis of requests with phatic communication]. Nihongo Kyooiku, 79 , 53–63. Klopf, D. (1991). Japanese communication practices: Recent comparative research. Communication Quarterly, 39 , 130–143. Koike, D. (1996). Transfer of pragmatic competence and suggestions in Spanish foreign language learning. In S. M. Gass & J. Neu (Eds.), Speech acts across cultures (pp. 257–281). Berlin: Mouton de Gruyter. Kubota, M. (1995). Teachability of conversational implicature to Japanese EFL learners. IRLT Bulletin, 9 , 35–67. Kusakabe, M. (1994). Japanese argumentation: A new perspective. Unpublished master’s thesis, University of Oregon, Eugene. Levinson, S. (1983). Pragmatics. Cambridge: Cambridge University Press. Makino, S., & Tsutsui, M. (1997). A dictionary of intermediate Japanese grammar . Tokyo: The Japan Times. Marriott, H. (1995). The acquisition of politeness patterns by exchange students in Japan. In B. Freed (Ed.), Second language acquisition in a study abroad context (pp. 197–224). Amsterdam: Benjamins. Maynard, S. K. (1993). Discourse modality: Subjectivity, emotion, and voice in the Japanese language. Amsterdam: Benjamins. Maynard, S. K. (1997). Japanese communication: Language and thought in context. Honolulu: University of Hawaii Press. Maynard, S. K. (1998). Principles of Japanese discourse. Cambridge: Cambridge University Press. McLaughlin, B. (1987). Theories of second language learning . London: Edward Arnold. Miike, Y. (2003). Japanese enryo-sasshi communication and the psychology of amae: Reconsideration and reconceptualization. Keio Communication Review, 25, 93–115. Miller, L. (1994). Japanese and American indirectness. Journal of Asian Pacific Communication, 5, 37–55. Miyaoka, Y., & Tamaoka, K. (2001). Use of Japanese language honorific expressions by native Chinese speakers. Psychologia, 44(3), 209–222. Mizuno, K. (1996). Interlanguage pragmatics of requests: Directness and perspectives. Papers on Language and Culture, 18, 57–71. Mizutani, N. (1985). Hanashi kotoba-no bunpou [Spoken grammar]. Tokyo: Kuroshio. Mizutani, O. (1981). The spoken language in Japanese life. Tokyo: Japan Times.

The Modern Language Journal 92 (2008) Mizutani, O., & Mizutani, N. (1987). How to be polite in Japanese. Tokyo: Japan Times. Mizutani, O., & Mizutani, N. (1988). Gaikokujin-no gimon-ni kotaeru nihongo note 1 & 2 [Japanese notebook for foreigners’ questions 192]. Tokyo: Japan Times. Morgan, J. (1978). Two types of convention in indirect speech acts. In P. Cole (Ed.), Syntax and semantics: Vol. 9. Pragmatics (pp. 261–280). New York: Academic Press. Nelson, G. L., Carson, J., Batal, M. A., & Bakary, W. E. (2002). Cross-cultural pragmatics: Strategy use in Egyptian Arabic and American English refusals. Applied Linguistics, 23, 163–189. Ohta, A. S. (2001). Second language acquisition processes in the classroom: Learning Japanese, Mahwah, NJ: Erlbaum. Okazaki, S. (1993). Stating opinions in Japanese: Listener-dependent strategies. In E. A. James (Ed.), Strategic interaction and Japanese acquisition (pp. 69–95). Washington, DC: Georgetown University Press. Osborne, J. (2000). Advantages of hierarchical linear modeling [Electronic version]. Practical Assessment, Research & Evaluation, 7 (1), 1–3. Retrieved March 10, 2008 from http://PAREonline.net/getvn.asp?v=7&n=1 Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Newbury Park, CA: Sage. Rose, K., & Kasper, G. (2002). Pragmatics in language teaching . Oxford: Oxford University Press. Ross, S. (1997). An introspective analysis of listener inferencing on a second language listening test. In G. Kasper & E. Kellerman (Eds.), Communication strategies: Psycholinguistic and sociolinguistic perspectives (pp. 216–237). New York: Longman. R¨over, C. (2005). Testing EFL pragmatics. Frankfurt, Germany: Gunter Narr. Rumelhart, D. E. (1993). Some problems with the notion of literal meanings. In A. Ortony (Ed.), Metaphor and thought (pp. 71–82). London: Cambridge University Press. Sacks, H., Schegloff, E., & Jefferson, G. (1974). A simplest systematics for the organization of turntaking for conversation. Language, 50, 696–735. Saito, H., & Beecken, M. (1997). An approach to instruction of pragmatic aspects: Implications of pragmatic transfer by American learners of Japanese. Modern Language Journal , 81, 363–377. Sameshima, S. (1998). The acquisition of fixed expressions and sentence-ending expressions by learners of Japanese. Journal of Japanese Language Teaching , 98, 73–84. Sawyer, M. (1992) The development of pragmatics in Japanese as a second language: The sentencefinal particle ne. In G. Kasper (Ed.), Pragmatics of Japanese as native and target language (pp. 83– 125). Honolulu, HI: Second Language Teaching and Curriculum Center. Segalowitz, N. (2000). Automaticity and attentional skill in fluent performance. In H. Riggenbach (Ed.),

575

Naoko Taguchi Perspectives on fluency (pp. 200–219). Ann Arbor: University of Michigan. Segalowitz, S. J., Segalowitz, N., & Wood, A. (1998). Assessing the development of second language word recognition. Applied Psycholinguistics, 19 , 53–67. Soler, E. M., & Flor, A. M. (2005). Pragmatics in instructed language learning. System, 33, 381– 546. Sperber, D., & Wilson, D. (1991). Loose talk. In S. Davis (Ed.), Pragmatics: A reader (pp. 540–550). Oxford: Oxford University Press. Sperber, D., & Wilson, D. (1995). Relevance: Communication and cognition (2nd ed.). Cambridge: Cambridge University Press. Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston: Allyn and Bacon. Taguchi, N. (2002). An application of relevance theory to the analysis of L2 interpretation processes: The comprehension of indirect replies. International Review of Applied Linguistics, 40, 151–176. Taguchi, N. (2005). Comprehending implied meaning in English as a second language. Modern Language Journal , 89 , 543–562. Taguchi, N. (2007). Development of speed and accuracy in pragmatic comprehension in English as a foreign language. TESOL Quarterly, 41, 313– 338. Taguchi, N. (2008a). Cognition, language contact, and the development of pragmatic comprehension in

English as a second language. Language Learning , 58, 33–71. Taguchi, N. (2008b). The role of learning context in the development of pragmatic comprehension: A comparison of gains between ESL and EFL learners. Studies in Second Language Acquisition, 30, 423–452. Takahashi, S., & Roitblat, H. L. (1994). Comprehension process of second language indirect requests. Applied Psycholinguistics, 15, 475–506. Takenoya, M. (2003). Terms of address in Japanese: An interlanguage pragmatics approach. Sapporo, Japan: Hokkaido University Press. Tateyama, Y. (2001). Explicit and implicit teaching of routines. In K. Rose & G. Kasper (Eds.), Pragmatics in language teaching (pp. 200–222). Oxford: Oxford University Press. Thomas, J. (1995). Meaning in interaction: An introduction to pragmatics. London: Longman. Verschueren, J. (1999). Understanding pragmatics. London: Arnold. Yoshimi, D. (1999). L1 language socialization as a variable in the use of ne by L2 learners of Japanese. Journal of Pragmatics, 31, 1513–1525. Yoshimi, D. (2001). Explicit instruction and JFL learners’ use of interactional discourse markers. In G. Kasper & K. Rose (Eds.), Pragmatics in language teaching (pp. 223–244). Cambridge: Cambridge University Press.

APPENDIX Sample J–PLT Implied Meaning Items with English Translations Indirect Refusal Smith: Satoo-san genki desu ka? Ima jikan arimasu ka? (Ms. Sato, how are you? Do you have time now?) Sato: Ah Sumisu-san, doo shitan desu ka? (Oh, Mr. Smith, what’s up?) Smith: Ee onegai ga arun desu ga. Kore nihongo no essei nandesu ga, boku no nihongo chekku shite moraemasen ka? (Well, I have a favor to ask you. This is an essay I wrote for the Japanese class. Could you please check my Japanese?) Sato: Korekara compyutaa no jugyoo ga hachiji made arun desu. (I have a computer class from now till eight.) Question: Which statement is correct? 1. 2. 3. 4.

The woman is going to check the man’s Japanese. The woman is taking a Japanese computer class. The woman can’t write Japanese essays. The woman can’t check the man’s Japanese now.

Conventional Indirect Opinion Smith: Satoo-san konnichiwa. (Hello, Ms. Sato.) Sato: Sumisu-san konnichiwa. Haruyasumi wa doo deshita ka? Dokoka ni iki mashita ka? (Hello, Mr. Smith. How was your spring break? Did you go somewhere?) Smith: Ee boku wa Frorida no hoo ni isshuukan hodo iki mashita. Otooto to imooto ga irun desu. (Yes, I went to Florida for a week. I have a brother and sister there.)

576

The Modern Language Journal 92 (2008)

Sato:

Aa soo desu ka. Yokatta desu ka, Frorida wa? (Oh, really. Was Florida good?) Smith: Boku wa Frorida wa amari. (For me, Florida is not very.) Question: Which statement is correct? 1. 2. 3. 4.

The man didn’t like Florida very much. The man enjoyed Florida very much. The man doesn’t know much about Florida. The man has an older sister in Florida.

Nonconventional Indirect Opinion Negative Opinion Sato:

Sumisu-san, genki desu ka? Ima nihongo no jugyoo desu ka? (Mr. Smith, how are you? Do you have a Japanese class now?) Smith: Hai kyoo mo kanji no testo ga arun desu. (Yes. I have another kanji test today.) Sato: Taihen desu ne. Tokorode senshuu no nihongo no kurasu no prezenteeshon wa doo deshita ka? Yokatta desu ka? (That’s too bad. By the way, how was your presentation last week for the Japanese class? Did it go well?) Smith: Sore wa mo owatta kara iidesu yo. (It’s over, so that’s all right.) Question: Which statement is correct? 1. 2. 3. 4.

The man’s presentation was not so good. The man’s presentation was very interesting. The man’s presentation is not over yet. The man’s presentation and quiz are on the same day.

Positive Opinion Sato:

Sumisu-san, senshuu wa hawai no koohii doomo arigatoo gozai mashita. (Mr. Smith, thank you very much for the coffee from Hawaii last week.) Smith: Doo itashimashite. Moo nonde mimashita ka? (You’re welcome. Have you tried it yet?) Sato: Hai kinoo asagohan no toki ni ruumumeeto to issho ni nomi mashita yo. (Yes, I had it for breakfast yesterday with my roommate.) Smith: Soo desu ka. Doo deshita ka, koohii wa? (Oh, OK. How did you like the coffee?) Sato: Mata kattekite kudasai ne. (Please get it for me again.) Question: Which statement is correct? 1. 2. 3. 4.

The woman had the coffee yesterday afternoon. The woman liked the coffee. The woman didn’t like the coffee very much. The woman wants to buy the coffee again.

Note. The conversations were played in Japanese. Directions and option sentences were given in English.