• No results found

When I was in Japan , I ate a lot offish.

3 There is a whole epistemological and methodological discussion on the why, how and where from of concepts (cf Bechtel 1988) It is not necessary to go into the details here; it is sufficient to point to the

3.4 Data elicitation methodology

3.4.1 Task type

All data collection sessions consist of communication tasks and free conversations between me (K.H.) and one informant, as this combination is considered most supportive of informants' oral production. In the following, the data collection methodology is presented in detail.

Conversation

Larsen-Freeman and Long (1991, 31) make the point that subjects' performance varies from task to task. It was attempted to cover this variation by encouraging free conversation, which allows for a diversity of topics and supports varied production.

An intuitive reason for the choice of free conversation as an elicitation device is that JSL learners initially feel insecure when talking in Japanese, and that they feel more relaxed if they can chose topics themselves. Ample experience shows that Japanese is a "difficult" language for a longer period than e.g. in English, and that second language learners feel that they cannot express themselves to any great extent. Therefore, it was important in the data collection sessions to encourage confidence and communication, and avoid communication breakdown. This was encouraged by leaving the choice of topic partially to the participants. Usually, the topics of conversation were weekend activities, cinema, holidays and the Japanese language course.

Another rationale for free conversation is that they are open-ended, so that the interviewer is free to ask for an infinite amount of further information. Pica et al. (1980) found that open-ended clarification requests elicit much interlanguage modification:

"Pica et al. (1980) found that open-ended clarification requests such as 'what?' or 'could you say that again?' led learners to modify their interlanguage significantly more often than the more containing confirmation checks which repeated the learner's production with rising intonation, e.g. 'books, did you say books? ' "

(Pica, Kanagy and Falodun 1993, 28)

Johnston (1985) observes that new structures initially co-occur with one or a few lexical items only, and are then incrementally used with a wider variety of words. It is therefore important to cover a wide range of topics and subtopics, so that the learner produces as many different lexical items and, perhaps, structures as possible. In conversations, the interviewer has the chance to steer the discourse and initiate new topics or a further elaboration of the current topic in direct reaction to the informant.

In short, conversation provides the researcher with the chance to influence the dialogue by prompting the informant, initiating new topics and offering the environment for a variety of structures, thereby supporting varied production.

Tasks

The choice of tasks for this study is based on research on communication tasks and the output they elicit (cf. Long 1983a, b, 1985a, b, Varonis and Gass 1985, Gass and Madden 1985, Doughty and Pica 1986, Duff 1993). Pica, Kanagy and Falodun (1993) summarise research of the field and propose a task typology which characterises different tasks according to the factors they found to be the most influential for the linguistic outcome.

Tasks are, in the current context, instruments designed to elicit production data (Larsen-Freeman and Long 1991, 27). Pica, Kanagy and Falodun (1993) define a task as a goal-oriented activity or work and make the point that

" ... a task is not an action carried out on task participants, rather a task is an activity which participants, themselves, must carry out."

(Pica, Kanagy and Falodun 1993,12)

Pica et al.'s (1993) task typology differentiates tasks according to the parameters of interactant relation, interaction requirement, goal orientation and outcome option (Pica et al. 1993,14f). They examine whether:

- the interaction goal is for the involved subjects identical, convergent or different; - subjects have identical or different information at their disposal;

- there is a one-way or two-way information flow; and - one or several outcomes are possible.

They conclude that tasks elicit the highest amount of interlanguage modification if interactants have different information, if a two-way information flow is required, if participants have same or convergent goals, and if there is only one acceptable outcome (Pica et al. 1993, 17). Learners should have to achieve a goal in cooperation, because then they are under pressure to make themselves understood and negotiate meaning where necessary. This leads to a "modification of interlanguage" (Pica et al. 1993,13), i.e. to richer data.

These parameters were taken into consideration when selecting tasks for the present study. The results for the implementation of specific factors, however, differ from Pica et al.'s parameters because most data collection sessions in this study are dyads between one informant and the researcher. Data from the researcher are not of interest, so a one-way flow of information from the informant to the researcher is desirable. The informant has to provide all information that is necessary to the goal of the task, which is the interlocutor understanding the informant's story.

The possible outcome of the task varies from session to session. There are picture stories that allow one interpretation only, and this restricts the possible outcome, which has the advantage that data from different informants are highly comparable. In other tasks, the material allows different interpretations. This has the advantage that informants can choose the interpretation they are able to describe, which helps to avoid a communication breakdown.

One further factor needs mentioning. In most studies on dyads, the interviewer- interlocutor is a native speaker of the target language in question. In the present study, the researcher who conducts the sessions is a second language learner of Japanese and has a first language (German) different to the informants' first language (English). This should not pose a problem, as Gass and Varonis' (1985) study on differences between NS/NNS- and NNS-NNS-dialogues found that more negotiation occurs when both participants are from different language backgrounds and are at different proficiency levels.

3.4.2 Data collection sessions

Data collection sessions were conducted at the end of each semester. "Session 1" is therefore the session at the end of the first semester, i.e. after three months of instruction, and "session 6" was conducted after the sixth semester. As the study progressed, the friendly acquaintance between the five informants and the researcher grew, and the informants developed a relaxed attitude to the sessions. From the learners' point of view, the sessions were part of their study routine, and they enjoyed the rare chance of being allowed to try out their Japanese for as long as they liked.

All communication tasks are based on visual aids like pictures and picture-stories. In the first session, two informants work together on a picture description and drawing- task. In all following sessions, informants are asked to describe picture-stories to their interlocutor.

The majority of the pictures that are used are from Rapid Profile, a tool designed to measure ESL development and define the developmental stage at which a speaker is (Pienemann 1992). However, the pictures themselves are designed to promote conversational interaction in English as well as other languages and are appropriate for all levels (Mackey 1994). The pictures used in this study are not necessarily used in the way suggested in the accompanying booklet (Mackey 1994), but for story retelling and as a starting point for further conversation on topics related to the stories.

First session

In the first data collection session, informants were asked to work in pairs and engage in an information gap task: a picture description. Each informant was given one picture and instructed to describe it to the partner, who then had to draw it. The information flow was one-way, but as informants swapped roles, data could be collected from the informants in both roles. Only one outcome was possible, which was expected to motivate negotiation between the informants. Results showed that in the main, the informant with the information produced more talk than the drawer. Most questions by the drawer were related to comprehension and only rarely demanded further information.

All pictures (see Appendix II - IV) were designed by the researcher except one (Appendix IV), which was part of the Rapid Profile package. Pictures contained only items for which the subjects knew the Japanese words. As a precautionary measure, and to create a more relaxed atmosphere, subjects were familiarised with a word list prior to the task.

After the tasks had been finished, the sessions continued in dyads between the researcher and one informant at a time. First, informants were asked to fulfil another picture description-task. In this task, the interviewer could motivate elaboration of utterances by asking questions. In the following conversation it was talked about what informants had done the evening or the weekend before, whether they did sport, worked part-time, etc. As this session was held after only three months of instruction, the informants felt insecure and easily frustrated and conversations were short.

Second session

This time, informants produced more and more varied data, as they had grown familiar with the situation and their interlocutor, were more relaxed and higher skilled in JSL. From the second session on, all data collection sessions were conducted between the researcher and one informant at a time.

The task in this session was the retelling of a picture story about a family-picnic. The pictures (see Appendix V) were part of the Rapid Profile package. Subjects were asked to put the six pictures that made up the story into a meaningful order and then tell the story. The fact that they had to order the pictures first gave them the chance to grow familiar with the contents. There was only one possible outcome. The one-way information flow of this set-up was the most productive, as it made the informant talk most. The researcher provided support where necessary and asked questions to elicit further information. The second data collection session also included free conversation. Informants talked about their weekend activities, study experience and other daily-life topics.

Third session

In the third round, another storytelling-task was set. Informants were provided with about 25 little cards, depicting everyday items including hees, a park, a bird, a dog, a bone, a restaurant, a supermarket, the ocean, a swimmer, a television set, a radio, etc. They were asked to choose as many of these pictures as they wished and make up a story with them. Again, this ensured that the informants were familiar with the story's contents.

This task was similar to the one in session 2, because it was another storytelling with a one-way information flow. The difference was that now the informants themselves defined the outcome of the story. In this session as well, questions were asked that provided an environment for different structures. The topics of the following free conversation covered again informants' every day life experience, holiday hips and films. Fourth session

The first task was the retelling of a picture story from the Rapid Profile-package, this time with a "surprise-effect", where learners found themselves in a situation where they had to find an explanation for how a big meal had suddenly disappeared out of a pan (see Appendix VI). The other task consisted of five pictures, each depicting an aspect of a police woman's job (see Appendix VII). Here, not a coherent story was to be described, but a typical working life, from paper work to arrests, had to be talked about. Informants played the role of the police woman and told the researcher about their work. This story, too, provided an opportunity for a wide range of questions by the interlocutor. Parameters of both tasks are identical to those described for sessions 2. The conversation covered the usual topics.

Fifth session

The fifth session included another Rapid-Profile picture-story that subjects were asked to retell; this time a little girl's visit to the circus (see Appendix VIII). This story was longer and more complex than the previous ones, and the interlocutor supported the informants by supplying vocabulary where necessary and by asking questions. Again, informants had to put the pictures into order and so defined the outcome themselves.

Informants were also asked to talk about a little fish that they saw on several pictures, each of which was only shown after the previous one had been described. In later pictures, the little fish was eaten by a bigger one (see Appendix IX). The idea for this task, which has been designed to elicit passive-constructions, was borrowed from Tomlin (1995); however, after two and a half years of instruction, and in spite of its having been taught for more than a year at that stage, only one student produced a passive-construction. This task formed only a small part of the session; the main parts consisted of the picture-description and a conversation on university, holidays, films, and books, among other topics.

Sixth session

Again, the main part of the session consisted of conversation. The main topics were jobs and plans for the following year, after the learners left university. A picture description task (see Appendix X) with the usual conditions was also part of the data collection.