(ENSs) data. Role play was used to elicit the EFL learners’ and the ENSs’ corpus.

Chapter Three Methodology

The purpose of this study is to investigate how well the college freshmen EFL learners in Taiwan can do in requests when compared to the English native speakers’

(ENSs) data. Role play was used to elicit the EFL learners’ and the ENSs’ corpus.

Based on the theoretical backgrounds discussed in the previous chapter, descriptions of participants (3.1), materials (3.2), procedures (3.3), coding schemes and measures (3.4) will be provided in this chapter.

3.1 Participants

In this study, we recruited two groups of subjects, namely, the English Native Speakers (ENSs), and the English-as-a-foreign-language (EFL) learners. There were six people in each group. These two groups of subjects were compared against each other to reveal how well the six GEPT intermediate-level EFL learners can do in requests.

All the EFL learners were male college freshmen

1

. They were all volunteers from a national university. Their mother tongue was Chinese. Students who had spent time in any English-speaking country for more than a month were excluded from this study. Three of these informants had never been abroad and three of them had but not longer than one month. To ensure the homogeneity of their proficiency level, we chose those who had passed the oral section of the intermediate level General English Proficiency Test

(GEPT) to be our participants. In addition, none of them passed the intermediate-high level in GEPT.

The average time for each group to finish the role play was 20 minutes. Each of them

1 Freshmen here refer to those who just graduated from high school for four or five months and thus were viewed as high school graduates.

2 High school graduates are supposed to pass the intermediate level of GEPT.

was ensured that all their information would only be used for the research purposes.

The six ENSs in the study all came from English-speaking countries. All of them were Americans except one, a Canadian. This, however, was not to influence our goal to observe the cultural differences between Chinese and Americans in their request uses, due to the geographical proximity and similar culture between America and Canada. All of them were in their twenties or early thirties. Four of them had bachelor degrees and two of them had MAs. They were all English teachers here in Taiwan except one who was an engineer. The average time for each group to finish the role play was also about 20 minutes.

3.2 Materials

In this section, we are going to discuss why we adopted role play as the elicitation method, and what situations were set up in the role play sheets.

3.2.1 Role Play as the Elicitation Method

Methods that have been employed for request productions include natural ethnographic studies (Ervin-Tripp, 1976), open role-play with interaction between players (Rintell, 1981; Walters, 1981; Trosborg, 1995), closed role-play with no interaction such as the Cartoon Oral Production Task (COPT) (Hsu, 2003), oral questionnaire (Hsu, 2001) and discourse completion test (DCT) (Blum-Kulka et al., 1989).

DCT

is often used to elicit linguistic strategies employed by the learners for specific speech acts (Rintell & Mitchell, 1989). Several strengths have been pointed out. For example, DCT can often elicit linguistic corpus from many subjects relatively easily at one time. In addition, it often contains controllable contextual variables in

3 In DCTs, participants are often provided dialogues with omitted turns in which they must fill for an aimed speech act (Blum-Kulka et al., 1989).

which the researchers are interested in (Rintell & Mitchell, 1989). Therefore, it is often considered as a useful tool especially in comparative studies.

However, there are many potential drawbacks. First, when compared with responses gathered from ethnography or interactive role play, responses in the DCT are often more confined because the subjects are usually only given one or two turns at most to respond (Trosborg, 1995). Secondly, the written corpus gathered by DCT often brings about two problems: formality and spelling (Rintell & Mitchell, 1989).

Written responses gathered from DCT are usually more formal and thus may be quite different from those in the spoken forms. In addition, the gap between corpus in DCT and those in interactive conversations can be quite wide owing to the absence of

“performance constraints”: the influence of affective factors associated with

face-to-face interlocutors, and time constraints (Trosborg, 1995, p. 305). Finally, it is doubtful as to how representative those corpus gathered by DCT really are of those in spontaneous speech (Rintell & Mithcell, 1989). Due to those aforementioned

weaknesses, DCT is criticized as only eliciting subjects’ pragmatic knowledge in interactions rather than their skills in actually applying the knowledge in real communication (Trosborg, 1995).

In contrast, in order to get a more complete picture of authentic speech,

researchers need to include contextual factors involved in authentic speech to reveal interpersonal and interactive features of speech acts (Wolfson et al., 1989).

Therefore, elicitation methods like ethnography or role play come into place.

Ethnography can often provide researchers with natural occurring speech acts along

with information about the interlocutors’ sex, age, status, culture or relationships by

means of observation or recordings. While it can often provide valuable insights into

how linguistic strategies are actually used in a given language and culture, some

drawbacks exist. First, with regard to contextual variables, ethnography is usually

quite time-consuming because the variables to be observed rarely emerge in the natural spontaneous speech within a short time. In addition, the contextual variables are often not replicable. Second, if it is gathered by observation, the researcher often needs to rely on his/her memory to correctly record the corpus. This heavy reliance on memory may distort the data to some extent (Wolfson et al., 1989).

Similar to ethnographic studies, role play can also provide researchers with speech act behaviors in its full discourse contents (Kasper & Dahl, 1991). In role play

, the situations are often given to the participants first and then the participants need to say what the person they are role playing would say in the situation. Several disadvantages have been pointed out. First, Rintell & Mitchell (1989, p.251) argue that since the participants are role playing and not “naturalistically engaged in the interactions under investigation,” we can often not be sure of the degree of the

authenticity of these oral productions when compared to those in spontaneous speech.

Secondly, a higher degree of anxiety can be induced by role play because subjects might think they were taking a test (Rintell & Michell, 1989). In addition, when the roles are quite distinct from the speaker’s real-life identity or unfamiliar to the speakers, it is doubtful that the valid responses would be obtained (Trosborg, 1995).

Despite these weaknesses, interactive/open role play is still chosen to be the elicitation method of the present study due to the following strengths. First, Rintell &

Mitchell (1989, p.251) maintain that unlike DCT, data gathered by role play can often be “a good indication of natural speech” because participants are not limited by the turns or spaces provided and can say whatever they feel like saying in a given situation. Secondly, role plays can examine not only linguistic responses but also the influence of contextual variables as stated by Rintell & Mitchell (1989, p.251):

4 The meaning of “role play” is defined as follows: “pretending to react as if one were someone else in a different situation” (see McDonough, 1981, p. 80, cited in Trosborg, 1995, p. 144).

“linguistic responses from role plays were proved to be sensitive to contextual variables under investigations (Rintell, 1979, 1981).” When compared to

ethnography, the advantages of role plays are that contextual variables of research value are usually replicable and controllable. Researchers usually do not have to spend so long a time for some particular contextual variables to emerge as in

ethnography. Thirdly, between the closed (Hsu, 2001, 2003) and the open forms

of role play, the latter one is even better in that it involves interactions among

interlocutors and requires active participation from the speakers in order to obtain the outlined goals (Trosborg, 1995). The many turns of interactions as well as active engagement required in open role plays make them the best representation of natural spontaneous speech. Therefore, we adopted the open/interactive form of role play as our elicitation method.

3.2.2 Content of Role Play Sheets

In this study, based on Byon’s (2004) design, a total of 6 request situations were created on our role play sheets

with three intervening ones, invitations (situation 3), apologies (situation 6) and gratitude (situation 9), inserted in between. The scenarios were described only in English for the ENSs. However, both Chinese and English were provided for the EFL learners to ensure that the subjects understood all the instructions clearly (appendix A).

Two social variables

, power and distance, which have been reported to be influential factors (Leech, 1983) in request studies, were included. Power contains

5 Celce-Murcia & Olshtain (2000) argues that interactive role plays in two languages often provide us with subtle linguistic and cultural differences in speech act behavior.

6 The roles in role play sheets are viewed as common figures that participants in our study would have had contact with or even familiar with by consultations with some graduate peers and ENSs.

7 In terms of the social variable of imposition, it is deemed as equal in each context through the consultation with several Taiwanese college students, graduate school students, and English-speaking foreigners during the process of finalizing the sheets.

three values (+, -, =) and distance only two (+ and - ). This results in 6 relationships or situations: unfamiliar professor [+power, + distance] (situation 1), familiar professor [+power,-distance] (situation 5), stranger [=power, + distance] (situation 2), best friend [=power, -distance](situation 7), junior club member [-power, + distance]

(situation 4), and younger roommate [-power, -distance] (situation 8), as shown below in Table 3-1.

Table 3-1: Social Constraints Embedded in the Role Play (Byon, 2004, p. 1677) Respondents relative social ranking with

A. unfamiliar group [+D] S1 Professor Higher [+P]

S2 Stranger Equal [=P]

S4 Junior club member Lower [-P ]

B. familiar group [-D] S5 Professor Higher [+P]

S7 Best friend Equal [=P]

S

Junior Roommate Lower [-P ]

The sheets which contained the six request situations together with the three distracting ones were distributed to the 12 participants. So a total of 36 intended request samples were collected from each group. Among each group, there were 12 [+P] situations, 12 [=P] and 12 [-P] while there were 18 [+D] and 18 [-D] scenarios respectively.

3.3 Procedures

The data collection lasted for two months. Before the study began, all

participants were asked to fill out a demographic survey on their backgrounds. The

issue of confidentiality of their data was ensured and the instructions were explained

by the researcher in the very beginning. They could ask questions related to the

situations or vocabulary before the role play began. No question was asked in the

recording process.

In the first month (i.e., the middle of November in 2004), informants in each group were randomly paired up; we had three dialogic pairs in each group, namely, the EFL, and the ENS groups. In each pair, one played the speaker role of making requests while the other played the role of the respondent. The speaker and respondent were given different role play sheets with the same contexts

but opposing roles (see appendix A). These talks were recorded.

In the second month (i.e., the middle of December), the participants reversed their roles and were paired up with those other than their original partners in the first month. Then they went through the whole process again. These three new pairs of dialogic data in each group were also recorded and transcribed. In the end, we had a total of 6 sets of elicited dialogic data in each group.

3.4 Coding Schemes and Measures

In this section, we will first discuss the major components of a request and their functions (3.4.1), ways of classifications (3.4.2), the coding reliability (3.4.3) and measures (3.4.4).

3.4.1 Segmentation

The segmentation of a request speech act in this study is primarily based on the coding manual of the Cross-Cultural Study of Speech Act Realization Patterns (Blum-Kulka et al., 1989), with revisions made to fit our data and research goals. As has been reviewed in the previous chapter, a request is composed of (1) an

obligatory request head act (RHA) with optional internal modifiers realized by lexical/phrasal downgraders and (2) optional external modifiers realized by

8 The speaker’s sheets were mainly the description of a given context which was identical with those of the respondent’s; however, the request task was only stated in the speaker’s sheet while the respondent was unwitting of the speaker’s intention.

mitigating supportive moves (RSMs). The segmentation

is illustrated as follows:

Situation: Asking the permission from a professor for taking a course.

S

I need to take your course but it’s closed./

(

I was wondering / if you can / kind of / help me out here/

(

(

because I really need this course here in order to graduate.

The function of optional modifiers, whether internal downgraders or external mitigating RSMs, is to decrease the impact of a face-threatening request and to make it less imposing or less direct and possibly more polite. This mitigating function is illustrated below, from the most direct one without any modifiers in (1) to the most indirect one heavily modified by internal downgraders and external RSMs in (5).

(1) Shut up! (RHA)

(2) Could you be quiet? (RHA) (3) It’s really noisy here (RHA).

(4) I was wondering

if you could turn down your voice a little bit?

(RHA)

(5) I get an exam tomorrow (RSM). I was wondering if you could turn down your voice a little bit? (RHA) I would really appreciate if you do that (RSM).

Examples (1)-(3) above demonstrate different levels of linguistic directness ranging from the most direct one in (1), a bare imperative, to conventionally indirect one in (2), with modal, past tense, and interrogative, and finally to the least direct one, the non-conventionally indirect request, namely, hints in (3). In example (4), the multi-lexical/phrasal downgraders (e.g., “I was wondering,” and “a little bit”) make the conventionally indirect request even less direct. Example (5) is considered

9 The model of classification is a revision, and expansion of Hsu’s framework (2003, p.47).

10 The italicized parts are lexical/phrasal downgraders.

the least direct request since not only conventionally indirect strategy but also multi-lexical/phrasal downgraders and RSMs are employed. Therefore, the illocutionary force of a request is often adjusted by the uses of internal lexical/

phrasal downgraders and external RSMs.

3.4.2 Taxonomy

In this section, we will focus on how we classified RHAs, internal modifiers, and RSMs into categories to meet the research goals of this study. In this study, request head acts are usually realized in different strategies. We combined

Blum-Kulka et al. (1989) CCSARP coding scheme with Zhang’s (1995b) three levels of linguistic directness: direct (D), conventionally indirect (CID) and

non-conventionally indirect (NCID) requests with some modifications to fit the current data. The result was a scheme with 6 categories

ranging from direct (mood derivable, explicit performative, want statement) to conventionally indirect

(suggestory, preparatory) and to non-conventionally indirect (hint), as shown in Table 3-2 below.

Table 3-2: The Modified Types of Request Strategies Directness Strategy Types Examples

1. Mood derivable Please, let me go to your class.

2. Explicit performative

I’ll just borrow your computer for a minute or so.

D

3. Want statement I’d like to audit your class that is coming up on Wednesday.

I need your computer, man.

11 For detailed definitions of each subcategory and more examples for strategies, please refer to appendix B.

4. Suggestory formula I think you can lend me your sunscreen after you use it.

CID

5. Preparatory Would you be able to direct me to the dormitory?

NCID 6.Hint I forgot my sunscreen lotion.

As for internal modifiers, they consist of both lexical/phrasal upgraders, downgraders and syntactic downgraders. However, due to lack of upgraders in our study, and the overlapping of lexical/phrasal downgraders and syntactic downgraders, our internal modifiers were confined to lexical/phrasal downgraders. They were lexical or phrasal choices used within the RHAs to mitigate the force of a request internally (Blum-Kulka et al., 1989).

To categorize the types of lexical/phrasal downgraders in our corpus, we

modified the CCSARP coding categories by Blum-Kulka et al. (1989) with references to Li (1989) and came up with a total of 6 categories

: politeness markers,

subjectivizers, consultative markers, conditionals, appealers, and downtoners, as shown in Table 3-3.

Table 3-3: The Modified Types of Lexical/Phrasal Downgraders Lexical/ phrasal downgrader

Types

Examples

1. Politeness marker Please let me go to your class.

2. Subjectivizer I was wondering if you’re done, maybe I can borrow it.

3.Consultative marker Do you think you can spare a few hours for me?

4.Conditional Can I get some of your sunscreen when you finish?

12 For detailed definitions of each subcategory and more examples for the lexical/phrasal downgraders, please refer to appendix C.

5. Appealer Just let me listen to your teaching, okay?

6.Downtoner I was wondering if you would kind of help me out here.

RSMs consist of both mitigating and aggravating RSMs

. Due to a lack of aggravating RSMs in our corpus, the RSMs in our study will be confined to

mitigating ones. We adopted the CCSARP coding scheme in Blum-kulka et al. (1989) with references to Schiffrin (1987), Li (1989), Redeker (1990) and Byon (2004). We categorized all the RSMs in our corpus into four major types: openers (i.e.,

identifications), grounders, imposition minimizers (i.e., preparators, mitigators, conditionals, appealers, downtoners, and discourse markers), and positive

reinforcement (i.e., rewards, compliments, gratitude, and good will). A total of 12 categories

of RSMs were established, as indicated in Table 3-4 below.

Table 3-4: The Modified Types of Request Supportive Moves (RSMs) RSM Types (Subtypes or sub-categories) /Examples

Opener 1. Identification

You might not know me but my name is Kevin.

Grounder It seems that I get a bit mixed up with the directions he gave me.

Would you be able to direct me to the dorm?

Imposition minimizer

1.Preparator

Would you like to help me out a little bit?

13 Aggravating RSMs include “insults, threats, and moralizing” (Blum-Kulka et al., 1989, p. 288).

14 For detailed definitions of each subcategory and more examples for the RSMs, please refer to appendix D.

2.Mitigator

It won’t take you too much time.

3.Conditional

If you let me listen to your teaching, I would very appreciate you.

4. Appealer

I need to audit your class this semester. Is that cool?

5. Downtoner

I’ll just get it down really quick.

6. Discourse marker

A. y’know: You know, Mary, black hair, kind of tall.

B. well: Well, I really need this course in order to graduate.

1.Reward

I’ll teach you some guitar skills but at first, I want you to give me some help.

2. Compliment

I know it’s cool and everything.

3. Gratitude

You’ll be really doing me a favor.

Positive

Reinforcement

4. Good will

If you need help, you can feel free to talk to me anytime

3.4.3 Coding Reliability

To make the coding more reliable, we asked two other graduate students in the English Department of NTNU to do the coding. They had to familiarize themselves with the coding schemes of the strategies, the lexical/phrasal downgraders, and RSMs before the coding process. After the researcher segmented the corpus into major categories for the request strategies, lexical/phrasal downgraders, and request supportive moves respectively, the other two graduate students helped to

sub-categorize all the corpus into strategy types, downgrader types, and RSM types,

based on the examples provided by the researcher. We resolved all the disagreeing

parts by discussions. The inter-rater reliability between the researcher and the other

two raters before the discussion was .98

. 3.4.4 Measures

Each move of the speech production related to requests was coded either into RHAs or RSMs. Subcategories were identified and their frequencies were counted.

Chi-square tests were conducted to evaluate if there was any significant difference

between the ENSs and the EFL groups in request strategies, lexical/phrasal

downgraders, and RSMs respectively in terms of general frequencies and frequencies under the influence of power and distance. The level of significance for the study was set at .05 for the P value.

15 The sum of request strategies, lexical/phrasal downgraders and RSMs are 391 counts. The problematic parts are 6 counts. Therefore, the agreement portion is about .98.

16 In the present study, significant deviations between the ENS and the EFL groups are determined by chi-square tests; however, the deviations in each sub-category are described in terms of tendency.