Discourse Variation in First Certificate in English

Discourse Variation in First Certificate in English Oral Interviews

A Report to the University of Cambridge local Examinations Syndicate

Richard Young Department of linguistics . Southern Illinois University at Carbondale

September, 1991

Table of Contents Introduction ............................................................................... I Discourse in Dyadic Interactions ............................................. 2 Interactional Contingency ................................................................ 2 Goal Orientation ............................................................................... 5 Dominance ....................................................................................... 9 The Discourse of Oral Proficiency Interviews ....................... Background to the Study ................................................................ Discourse Variables ........................................................................ The Influence of Context ................................................................

11 11 12 15

Analysis and Results ................................................................ 18 Discussion ................................................................................ 25 Contextual Influences on the Candidate .......................................... 26 Contextual Influences on the Examiner .......................................... 26 Contextual Influences on the Discourse as a Whole ............ ,........... 27

Conclusions .............................................................................. 29 Recommendations for Further Research and Development 31 Further Development of the Model ................................................. 31 Comparative Analysis of Other Discourse Genres .......................... 32 Further Investigation of Task, Theme, and Interlocutor .................. 32 Comparative Analysis of Other Interview Formats ......................... 33 Oral Interview Design Considerations ............................................ 34

Notes ........................................................................................ 36 References ............................................................................... 3 8

Introduction

In recent years much effort has gone into the study of the speech of second language learners in different social contexts and into relating the variation in interlanguage phonology, morphology, and syntax observed in learners' speech to certain features of context'. It has been established that much of the variation in learners' spoken language is rule governed and relates in a systematic way to contextual factors, including the particular task in which a learner is engaged, the interlocutor or interlocutors with whom a learner is interacting, the topic or theme that the learner is talking about, and the linguistic environment of the form2 • These studies of phonological, morphological, and syntactic variation in interlanguage speech lead us to expect that features of interlanguage spoken discourse may be similarly sensitive to the context in which the discourse is constructed and pioneering work in describing variation in interlanguage discourse has indicated that this is, indeed, the case (Takahashi, 1989; Woken & Swales, 1989; Zuengler, l 989a, b). The present study contributes to this developing research tradition with an investigation of the possible influence on the discourse of oral proficiency interviews of the four variables of interlocutor, theme, task, and gender. In addition to the insights that it may bring into the structure ofinterlanguage discourse, the investigation of variation in oral proficiency interyiews is also an important practical undertaking. As Tarone ( 1988) has pointed out, the abundant evidence linking interlanguage variation to features of the context in which it is elicited has important implications for the design of instruments that attempt to test and evaluate learner's proficiency in a second language. In evaluating learners' proficiency, developers and users of language tests in effect make generalizations from a sample ofinterlanguage elicited in one context to learners' petformance in other and different contexts. What we now know about the systematic influence of context on learner's language may give us pause before we generalize from learners' performance elicited in one context to performance in others. ·

Discourse in Dyadic Interactions This report will concentrate on describing three aspects of the discourse of oral proficiency discourse: interactional contingency, goal orientation, and dominance. These three aspects have been discussed at length in recent research on interactions between native and nonnative speakers ofEnglish (van Lier, 1989; Zuengler l 989a, b). These three aspects have been chosen because they appear to be the most promising in elucidating the discourse structure of oral proficiency interviews and because they may help us to make a principled comparison between oral interviews and the discourse created in other forms of NS-NNS interactions.

lnteractional Contingency The discourse structure of oral proficiency interviews has been discussed in detail by van Lier (1989), who underscores the differences between the discourse of oral interviews and conversation. From the theoretical perspective of conversation analysis, van Lier proposes a model of dyadic interaction based on the work of Jones and Gerard (1967). This model describes dyadic interaction in terms of two dimensions: contingency and goal orientation. Contingency is the social detc1minant of the structure of interactions: that is, how participants react to each other, while goal orientation reflects the internal goals of each speaker separately3 • The question of how to apply the concepts of contingency and goal orientation in the Jones and Gerard model to the linguistic dimension of interaction is not a simple one. Jones and Gerard are social psychologists and their model was originally intended to explain social behaviors such as influence and attitude change. Although the model has been adopted as an explanatory construct by discourse analysts such as Clarke and Argyle ( 1982) and by van Lier (1989), as far as we are aware no previous attempt has been made to apply the model to an empirical analysis of discourse. The model promises to help in describing and explaining the dynamic structure of dyadic discourse since contingency and goal orientation may be seen as ways of describing local and global discourse dynamics. Contingency is a way of looking at the local dynamic ofhow participants create shared meanings in any one exchange, while goal orientation focuses on how those meanings evolve and change over longer stretches of discourse. · In this report the social-psychological model of Jones and Gerard has been na1TOwed by defining contingency as a property of sequences of speech between two interactants. In this version of the model, a contingent utterance is one in which the content and often the form of the utterance depend in some way on a previous utterance. When the utterance is contingent upon a previous utterance by the other participant, we describe the relationship between the two utterances 2

•.. •II II II

•· •Ill

DISCOURSE IN DYADIC INTERACTIONS

as "reactiveness." The most obvious example ofreactively contingent utterances are the familiar adjacency pairs (Schegloff, 1978) such as the question-answer sequence in this example4 • NS: NNS:

II II

NNS: NS: NNS: NS: NNS:

II

'

How long is the training to be a teacher for secondary school? How long? Is the training how long does it take to become a teacher Em for my school? Mm

Em I I 'ave to eh xxx eh course for eh four years

Another kind ofcontingency occurs when one participant's topic is taken up and developed by the other participant in the interaction as in the next example. In this dialogue, the NNS 's topic of being a musician is ratified and expanded upon by his native speaking interlocutor. NNS:

II II

Must be classic music because eh . is piano and flute and eh kind of classic instruments

Contingency is also evident in such negotiation ofmeaning sequences as those illustrated below. In this case, the non-native speaker's (NNS's) confirmation check, how long? and clarification request.for my school?both represent attempts to establish a common topic for the discourse. The underlying assumption of participants in such negotiation ofmeaning sequences is thatthey should be talking about the same topic, even though the NNS is unsure what that topic is. NS:

II

Have you any idea what kind of music it might be?

NS: NNS: NS:

So sometimes uh ththere are some problem like ththere are some problem like that . but uh . this uh it is something that I like very much so no one can no no one and nothing can stop me Yes well I think that is one of the . prerequisites of being a good musician lthat it is something you must= I IOh yes s' passion I =love and have passion for exactly

However, as has been pointed out by conversational analysts, some utterances appear to be unrelated in te1ms of topic but are nonetheless contingent because of an implied proposition understood by both parties which links the two utterances. This is exemplified by the following example in which the NNS' s utterance, I am 3


a swimming teacher, is understood by both parties to be contingent on the native speaker's (NS' s) preceding utterance, I've never had lessons. The two utterances are linked by the pragmatic implication that a swimming teacher gives swimming lessons. NNS: NS: NNS:

I

NS:

NNS: NS:

Uhm • can you swim?

Not very well Mm . why not . didn't you . uh take a course or didn't lyou have lessons?! I I've never I I've never taken a course I've never had lessons I started llearn- I II I am a swimning teacher (laughs) Really?

Although in social interaction, contingency is always present, it may be realized only by regular patterns of turn taking, prosodic features of participants' speech, or by non-verbal means such as gaze and posture. However, on the purely linguistic level there may be no apparent semantic relation between participants' adjacent utterances. This property of conversational sequences is referred to here as disjunctive contingency. Disjunctive contingency may be accompanied by lexical boundary markers in the discourse such as thankyou and excuse me in this example: NNS:

NS: NNS: NS:

But I: . the music is uh is il1J?ortant things of the life I think uh that uh every person um has uh personal relation with music and em and . and uh em everyone uh em has uh an i!1J?ortant uh an i!1J?ortant relation with music and uh it uh will probably remember uh remind uh uh remind him uh uh uh some particular set of his life and em . the music uh there are uh . infinity types of music and uh Thank you thank you Yes Eh excuse me . I'd like you to look for a minute at this list of skills

Alternatively, disjunctive contingency may be signalled by contra~tive stress as in: NNS: NS:

4

No . er I play sport yes (cough) . sometimes I swim play tennis Yes yes you don't Jmij; I imagine?

• •Ill •I • •II • '• II •II ' • ' .


When disjunctive contingency is not overtly signalled in the discourse, rapid shifts from one topic to another may cause comprehension problems for theNNS, as is illustrated in the next extract. In this dialogue, the shift by the NS from the first topic, school, to the second topic, this question, referring to a printed question on a handout, is not followed by the NNS, who continues talking about the first topic. NS: NNS:

NS: NNS:

Okay . are you still at school? Yes Eh what about this question here Ein it's . . it isn't very good because ha I don't know yet what I do but . I it's a good school

As illustrated in the above extracts, contingency is a property of adjacent turns in dialogue in which the topic of the preceding tum is coreferential with the topic of the following tum. In two special cases-negotiation of meaning and pragmatic implication--coreference of topics in adjacent turns is established by recourse to an underlying and unstated proposition linking the two. When contingency on the surface linguistic level is lacking, the disjunction is often marked in some overt way but where disjunctive contingency is not marked, the discourse often becomes difficult to interpret for the NNS 5 • These two properties of disjunctive utterances suggest that speakers expect contingency in dyadic interaction.

Goal Orientation The other dimension of the Jones and Gerard model represents the speakers' attempt to realize certain internal goals or plans through the interaction. As Jones and Gerard (1967) point out, "although normally these plans are vague and implicit, ... a plan tends to become quite prominent in awareness when important goals are at stake" (1967, p.. 506). In the context of linguistic interaction, goal orientation is evident in the discourse when topics persist over a large number of turns, or when a speaker returns to a topic raised earlier after a number of intervening turns on a different topic. Goal orientation is particularly evident in dyadic interactions in cases when one participant is following an agenda established in advance. In the oral proficiency interviews which form the object of the present study this is evident when the native speaking examiner follows the procedure established for such interviews by the examination board. For example, the instructions to oral examiners state, "Indicate to candidate(s) selected passages( s). Ask candidate(s) to skim passage before relating it to theme/ photographs" (University of Cambridge Local Examinations Syndicate, 1990,p.

s

J

! DISCOURSE IN DYADIC INTERACTIONS

3). And theresultin the extract below is thatthetopic of the threepassages persists over 12 turns in the interview-indicating that the topic was planned and that it is viewed as important by the NS for the successful outcome of the interaction. NS: NNS: NS: NNS: NS: NNS: NS: NNS: NS: NNS: NS: NNS:

Good look at the three passages. on the second page Eunice . and choose the one which fits your picture Ah choose the one which fiMhm

The third one Why is that? Well well because the other . the other two are talking about the music= Yes =and the other it's talking about the . building school I don't know 'Ihat' s right yes And so that . the third one that is the correct Yes It's the timetable of the classes

=

On the other hand, lack of goal orientation is evident in the rapid decay oftopics over one or two turns. In the oral proficiency interviews discussed here, the NNS candidate does not share the examiner's agenda, and ifthe candidate ha~ prepared any plans ahead of the interview, these may not be realized because the candidate does not control the introduction of topics. An example of this low topic persistence is the rapid topic shift from salary to holidays to colleagues at work in this final example. NS: NNS: NS: NNS: NS: NNS: NS:

Em wha- what things are most important to you when you're thinking about a job Well a good salary is the first and most important the salary and ... What else what other inportant factors are there to think about No I think I wouldn't mind the rest if I take a good salary No holidays Ah I wouldn't mind (laughs) What about horrible friends eh colleagues at work

Jones and Gerard combine the two dimensions of contingency and goal orientation in four styles of dyadic interaction. Pseudocontingency is characterized by a high degree of goal orientation but little reactiveness, asymmetrical 6

• DISCOURSE IN DYADIC INTERACTIONS

'I • •I I II

• • •II II II

II II

Figure i, Styles of dyadic discourse related to contingency and orientation of speakers to Internal goals. Based on Jones & Gerard (1967, p. 507).

A

B

R i~R Rti i i~R Rtf i i~R Rti i ~R

Pseudocontingency

A

B

A

B

A

B

R R 'R i'R ~'R i'R Rti~ R/~ R/i

•'R R(~

~'R

R/~

•'R R/.

Asymmetrical contingency

Reactive contingency

· Mutual contingency

i'R i efR V'R R ~ R/~ R/i R 'R 'R

contingency is characterized by a high degree of goal orientation by one party and a high degree ofreactiveness by the other party, reactive contingency is characte1ized by little goal orientation by either party but a high degree of reactiveness by both, and mutual contingency is identified as a high degree of goal orientation and reactiveness by both parties. These four styles are represented graphically in Figure l, in which solid arrows represent the predominant or major source of influence on each tum (R) by participant A or B and unfilled arrows represent the less important or minor source. The figure represents the flow of time from top to bottom, with A's turns on the left and B's turns on the right. Pseudocontingency represents the kind of dyadic interaction which each participant pursues his or herownintemal goals but theirreactiveness to each other is only conditioned by the requirement for orderly tum taking. Two actors in a play, or ritual social interaction such as weddings and inaugurals exemplify pseudocontingency.

in

7


Asymmetrical contingency represents the kind ofinteraction in which only one of the two participants has established goals for the interaction. The utterances of the other party are contingent upon those of the goal-oriented member of the dyad. There is, however, little reciprocal contingency on the part of the goaloriented member, whose behavior is largely driven by internal goals. Van Lier and Jones and Gerard suggest that such asymmetrical contingency is a characteristic of the discourse of interviews conducted by trained interviewers. Van Lier contrasts the asymmetrical contingency ofinterviews with the greater degree of reactiveness evident in conversations and in serious discussions and negotiations. In terms of the Jones and Gerard model, a conversation which switches topic freely withoutthe goals ofany one party dominating the interaction illustrates reactive contingency. Greater goal orientation by both par.ties in the interaction together with a high degree of reactiveness in both directions is characteristic ofmutual contingency. Such an interactional structure may be seen in serious discussions and negotiations. The Jones and Gerard model of dyadic interaction is suggestive of differences to be found between different styles of face-to-face interaction in dyads. The model itself is clearly theoretical rather than empirical, broad brush rather than finely detailed. How any given stretch of actual discourse fits into the model is open to question. It may be difficult to distinguish in practice between asymmetrical and reactive contingencies because some parts of the discourse correspond better to the asymmetrical model while other parts correspond to the reactive model. We have also seen that there exist different kinds of contingency in discourse such as question-answer, negotiation of meaning, and topic development. The picture is not so simple as the solid and broken lines in the Jones and Gerard figures suggest. Nonetheless, this model does provide a theoretical framework in which to conduct empirical studies of dyadic interaction involving second language learners. In particular, as van Lier has suggested, the distinction between asymmetrical contingency on the one hand and reactive and mutual contingencies on the other suggests ways in which the discourse of oral proficiency interviews may be compared with other forms of oral interactive discourse. In conducting thepresentstudytheJonesandGerardmodelhasbeenoperationalizedandrefined with the aim of increasing our understanding of dyadic linguistic interaction in general and the discourse which is constructed between native and nonnative speakers of English in oral proficiency interviews in particular. I

.i

:

8

• •II •

•II • •II • • •II

• '• •


Dominance Dominance is a further dimension of dyadic interactions which has received considerable attention from second language acquisition (SLA) researchers. Dominance in dyadic conversational interaction may be defined as the tendency for one participant to control the discourse by various means, including holding the floor, initiating and terminating topics, and controlling the other participant's access to the floor by means ofinterruptions and questions. In Levinson' s ( 1991) · terms, conversational dominance by one participant in a dyad limits the other participant's right to speak. In research on NS-NNS interactions, dominance by either party has been shown to be a function of the linguistic proficiency of the NNS, the relative expertise of the two participants in the subject matter of the conversation, and the gender of the participants. As Woken and Swales (1989) suggest, the dominance of NSs which was reported in many early studies ofNS-NNS interaction is probably due, at least in part, to the limited linguistic proficiency of the NNSs in those studies. However, more recent studies involving NNSs with higher ESL proficiency such as international graduate students at U.S. universities have shown that, at higher proficiency levels, the question of which participant dominates in NS-NNS interactions is more complex. Studies by Zuenglerand by Woken and Swales have shown that when NNSs either possess greater expertise in the subject matter being discussed or when both participants are led by the experimenter to believe so, then NNSs tend to dominate the interaction by speaking more and by "i;noving the task along" (Zuengler, 1989a) , namely, terminating discussion on one topic and initiating discussion on another within the same general thematic area. It has been suggested in a study of interactions between native speakers, that expertise and gender interact to produce dominance by one of the conversational partners (Leet-Pellegrini, 1980). Leet-Pellegrini found that gender alone did not account differences in dominance in her study of dyads, but that male experts were significantly more talkative than nonexperts or female experts. As far as genderrelated dominance in NNS-NNS interactions is concerned, Varonis and Gass (1986) found that males talked more than females in opposite gender dyads. The only studies of interactions between native and nonnative speakers to reveal possible gender-related differences are two studies of same-gender dyads by Zuengler. In a study of 45 NS-NNS female dyads, Zuengler (1989a) found that the NNS women interrupted their NS interlocutors more often when they were led to believe that they had superior expertise in the subject being discussed-art

9


appreciation. However, in a further study of27 NS-NNS male dyads, Zuengler ( l 989b) was unable to find any difference in interruption behavior attributable to differing degrees of expertise. Differences in patterns of conversational dominance which are attributable to gender suggest that other culturally relevant factors may also play a part in determining dominance, including socioeconomic status, which is currently being investigated by Levinson(1991 ). These results also suggestthatdominance is not a unitary construct, but rather a tendency that may be realized in different ways by different features of discourse. To summarize this review of previous research into the discourse ofNS-NNS dyadic interactions, we have seen that this may be characterized by (a) the semantic relation between one participant's utterance and the immediately preceding utterance by the other participant (interactional contingency), (b) the degree to which participants in an interaction are goal-oriented as indicated by the persistence oftopics in the discourse, and (c) the degree to which eitherparticipant dominates the discourse by gaining, holding, or ceding the floor, and by controlling the other participant's access to the floor through questions and interruptions. We have also seen that the organization of discourse created ir. dyadic interaction is related to the situational context of the interaction. Different conversational genres such as ritual interactions, interviews, casual conversations, and serious negotiations (Neu, 1986) may produce different structures of discourse. Similarly, the identities and roles ofparticipants, as native or nonnative speakers, as experts or nonexperts, as men or women also affect the structure of the discourse. We turn now to discuss an empirical study of the discourse of one kind ofNS-NNS interaction-oral proficiency interviews-and how it relates to the factors we have outlined above.

I

!

10

)

j i

~

'

1

j

• • • •I •II •I I

• • • • • ' II •II

The Discourse of Oral Proficiency . Interviews The study that will be described in the remainder of this report is motivated by two concerns. The first aim of the study is to provide a description of the discourse of oral proficiency interviews in terms of the three dimensions of discourse structure discussed above--interactional contingency, goal orientation, and dominance. This description will enable us to make a principled comparison between oral proficiency interview discourse and other genres ofspoken NS-NNS interaction. In particular, we wish to compare the discourse constructed in oral proficiency interviews with conversational discourse constructed by native and nonnative speakers in non-interview situations. The second aim of the study is to relate variation in the structure of discourse in oral proficiency interviews to features of context. To this end, we will attempt to relate discourse structure to the task in which NS examiner and NNS candidate are engaged, to the theme or subject matter of the interview, to individual differences among NS examiners, and to the gender of the examiner and candidate. If we understand which contextual factors contribute.most to determining the structure o'fthe interview, we may be able to use this information to design more valid instruments for assessing oral proficiency in the future .

Background to the Study The study is based on oral interviews of the First Certificate in English examination (FCE) of the University of Cambridge Local Examinations Syndicate (UCLES). The FCE is administered in about 70 countries world wide and is taken by approximately 180,000 candidates each year. The present interview was designed in 1984 and is a highly structured fonnat involving three separate tasks to be performed in a fixed order (UCLES, 1982). Different packs of thematically related materials are made available to examiners, who may choose which pack of materials to use for a given interview. The interviews generally last 12-15 minutes. The interview format allows several combinations of candidates and examiners to participate in one interview, including two candidates-ne examiner, two candidates-two examiners, and three candidates-two examiners. However, the most frequent arrangement is one candidate-ne examiner and in this study only the one-to-one combination was considered in order to make for more effective application of the model of dyadic interaction proposed and to facilitate comparison with other studies of dyads. The data for this study consist of 30 oral interviews, 14 recorded at one testing center in Rio de Janeiro, Brazil, and a further 16 in centers at Rome and Trieste in Italy. Recording equipment was a small tape recorder controlled by the examiner and in full view of the candidate. All tapes were transcribed by native speakers of British English in Cambridge and checked by the author. Transcripts 11

THE DISCOURSE OF ORAL PROFICIENCY INTERVIEWS

were coded by the author and a native speaking research assistant in Carbondale. All examiners in the study were NSs of British English who had experience interacting with students of English. The examiners both conduct the interview and score the candidate's performance on six five-point scales of fluency, grammatical accuracy, sentence pronunciation, the pronunciation of individual sounds, interactive communication, and vocabulary resource. The candidate's final score is the sum of the scores on the six separate scales. There were a total of eight examiners involved, of whom five were males and three were females. All examiners conducted at least two interviews, some conducted more. Candidates in Brazil were all native speakers of Brazilian Portuguese, while those in Italy were all native speakers of Italian. Candidates' spoken English proficiency as measured by the FCE oral interview ranged from intermediate to advanced. Seven interviews were conducted between female examiners and female candidates, five between female examiners and male candidates, 11 between male examiners and female candidates, and seven between males. In 12 interviews examiners chose the pack of interaction materials focusing on the general theme of" Learning," seven interviews focused on the theme of "Having a Good Time," and 11 focused on "Work." Each interview consisted ofa warm up, three separate tasks, and a closing. The wa1mup and closing werenotincluded in the analysis and the data consisted solely of the discourse in the three tasks. These were always performed in the same order and consisted of (a) a discussion based on one or more photographs, which we refer to as the photographs task, (b) a task in which candidates were asked to relate a p1inted passage to the photographs they had discussed in the photographs task, which we refer to as the passages task, and (c) an activity which' involved candidates expressing personal preferences about items in a list of activities related to the theme of the interview. This task we refer to as ranking activities.

Discourse Variables In the present study the basic unit of discourse chosen for analysis is topic6 • Topic, defined as the person or thing about which something is said, is a suitable discourse construct for analyzing contingency and goal orientation in dyadic interactions since it is possible to identify whether one participant continues or terminates a topic introduced by the other participant. Previous empirical studies of topic continuity in spoken discourse (Givan, 1984) and in written discourse (Schneider & Connor, 1990) have developed both the theory of topic continuity and have gone to considerable lengths to describe how, in practice, to identify coreferential topics in succeeding utterances.

' ' I

12


• • • •II •i • • • • ' '• II

It should be noted, however, that there are several limitations in using topic as

the basic unit of analysis. First, clearly, participants do not react solely to what their interlocutors say. They may also react in terms of eye contact, posture, and body positioning vis-a-vis their interlocutor as Argyle and Dean (1965) and Erickson and Shultz (1982) have shown. Similarly, as other analysts offace-tofaceinteractionsuchasGumperz(l979)haverevealed,participantsmayalsoreact not only to what theirinterlocutors say, but to how they say it-in particular to the intonation ofutterances. A full analysis of contingency in dyadic interactions will therefore have to deal with more dimensions ofthe interaction than are considered here. Second, as Brown and Yule (1983) and Giv6n (1984) have indicated, topic is not an either/or property of constituents of utterances. Rather, the many constituents of an utterance which are all to a certain extent "given" rather than "new," "topic" rather than "comment" suggest that with topic one is not dealing with an either/or property of discourse but rather a gradable one. Moreover, even if a topic can be identified, what is the scope of the topic? To address these problems, wechoseto adaptthetopical structure analysis ofSchneider and Connor (1990) to spoken data, using the t-unit as the stretch of discourse over which a topic has scope. The discourse variables measured in this study are defined in Figure 2. Dominance was measured in two ways: by comparing the quantity of each participants' stalk and the number of topics initiated by each participant. EQ, the quantity int-units ofthe examiner's talk during one task was measured, as was CQ, the quantity int-units of the candidate's talk during one task. However, given that the aim of these interviews is to elicit as large a sample of spoken English from the candidate as possible, then clearly a greater value of CQ than EQ does not necessarily indicate that the candidate is dominant as it may do in other genres of spoken interaction. Perhaps a better measure of dominance in the context of these interviews is the number of topics initiated by each participant. This measure is similar but broader than the number of questions asked by each participant, which has been used by previous researchers as a measure of dominance. ET, the number of topics initiated by the examiner in one task was measured, as was CT, a similar measure of candidate topic initiations. As will be seen below, the measures of numbers of topic initiations are also useful in characterizing the type of contingency of the interaction. Measures of contingency were ETR, the percentage of topics initiated by the examiner during one task which are subsequently taken up, or ratified, by the candidate, and CTR, the percentage of topics initiated by the candidate during one task which are subsequently ratified by the examiner. Goal orientation was measured by the persistence oftopics in the discourse, that is, the number oft-units IJ


Figure 2. Discourse variables

I

•

Measures of dominance EQ Quantity of the examiner's talk during one task int-units CQ Quantity of the candidate's talk during one task int-units Measures of dominance and contingency ET Number of topics initiated by the examiner during one task CT Number of topics initiated by the candidate during one task Measures of contingency ETR Percentage of topics initiated by the examiner during one task which are subsequently ratified by the candidate CTR Percentage of topics initiated by the candidate during one task which are subsequently ratified by the examiner

Measures of goal orientation ETP Mean persistence (int-units) of topics initiated by the examiner during one task CTP Mean persistence (int-units) of topics initiated by the candidate during one task in the discourse of either speaker between the first and the last mention of a topic. Again, this measure was calculated for both speakers: ETPisthemean persistence int-units of topics initiated by the examiner during one task, and CTP is the mean persistence in !-units of topics initiated by the candidate during one task. The distinction made in Schneider and Connor's work between parallel topic progressions (those which involve topic continuity in immediately succeeding t-units), and extended parallel topic progressions (those which involve returning to a previous topic afterone or more !-units containing a different topic) was not made in this study. The only exception was the arbitrary one that when a previous topic was returned to after 20 or more t-units, it was not coindexed with the previous topic but was considered to be a new topic. Given these measures, one can perhaps distinguish the fourinteractional styles proposed by Jones and Gerard as follows. Asymmetrical contingency involves more topic initiations by one party than the other and also greater ratiffcation of topics by one party than the other. If FCE oral interviews exhibit asymmetrical contingency, then ET will be greater than CT and ETR will be greater than CTR. The distinction between reactive and mutual contingencies in the Jones and Gerard model has to do with the degree of goal orientation involved. Reactive contingency involves fairly low topic persistence whereas mutual contingency involves high topic persistence as both parties pursue their internal goals. 14


• •Ill •II • • • • • • •II •I II

Reactiveness is high in both cases. Thus, if the interviews exhibit reactive contingency, we are likely to see a large number oftopics ratified by the other party (high ETR and CTR), however CTP and ETP will be low with neither participant predominating. On the other hand if the interviews exhibit mutual contingency, we are likely to see both high ETR and CTR as well as high ETP and CTP. On the basis of the above discussion and the previous research into NS-NNS discourse summarized above, the following hypotheses were constructed regarding the structure of discourse in the FCE oral interviews. HI.

Since the purpose of the oral Interview is to elicit a ratable sample of speech from candidates, candidates will talk more than examiners, i.e., CQ will be greater than f.Q .

H2.

Examiners will dominate the discourse by initiating more topics than candidates, i.e., ET will be greater than CT•

HJ.

Oral interviews will exhibit asymmetrical contingency which will be evident because of greater reactiveness by candidates and greater orientation to goals by examiners. Specifically, HJa.

Candidates will ratify a greater proportion of topics Initiated by their interlocutor than will examiners, I.e., ETR will be greater than CTR •

HJb.

The persistence of topics initiated by examiners will be greater than the persistence of topics Initiated by candidates, I.e., ETP will be greater than CTP.,

The Influence of Context The second research question concerns the relationship between discourse in oral proficiencyinterviews and the social context ofthose interviews, in particular, who the candidates are speaking to, what they are speaking about, and what task they are engaged in. The hypotheses are derived from previous findings on variation in phonology, morphology, and syntax in interlanguage speech, which we believe may be equally apparent on the discourse level. Several studies of interlanguage phonology designed within the general framework of speech accommodation theory have found that the learner's interlocutor plays a significant role in conditioning variation in the learner's second language pronunciation. Beebe ( 1977) and Beebe and Zuengler ( 1983) attributed this interlocutor effect to ethnic identification between participants. Berkowitz ( 1989) found a similar effect and attributed it to the degree of cultural 15


i'.,

I

empathy between NS and NNS. Selinker and Douglas (1985) claim to have observed an interlocutor effect in interlanguage discourse in NS-NNS conversations, but this was inseparable in their study from the theme of the conversation. Other studies, however, have cast some doubt on whether accommodation to the speech patterns of an interlocutor is as widespread a phenomenon in other areas of IL as speech accommodation theorists have claimed (Young, 1986, 1991; Zuengler, 1991). In theFCE oral interviews the NNS's interlocutor is clearly the NS examiner. Hypothesis H4 was formulated to see whether the discourse structure was related in some way to individual differences among examiners. H4.

I.

The theme (or topic, see note 2) ofNS-NNS conversations has been shown to affect learners' pronunciation by Eisenstein and Starbuck (1989), whp attribute the effect to learners' emotional investment in the theme of the conversation. Similarly, learners' expertise on a given theme has been shown to affect IL discourse by Selinker and Douglas (1985) and by Zuengler (1989a). Again, however, results in this area are not unequivocal. Smith ( 1989), for example, found that learners' comprehensibility was not different when they were speaking on general themes 01: when they were speaking on themes within their academic major. In the present study, the theme of the FCE oral interviews was identified as one of the three thematic areas of "Learning," "Having a Good Time," and "Work" according to the pack of materials chosen by the examiner. Would the selection of one theme over others affect the discourse, possibly because of candidates' expertise or emotional investment in one or more of these areas? This question was formulated as hypothesis H5.

11

:Ii ifi

( !

:1 I

Dominance, contingency and goal orientation In the discourse of FCE oral Interviews (as measured by quantity of talk, number of topic Initiations, reactiveness, and topic persistence) will vary according to the examiner.

i

HS.

Dominance, contingency, and goal orientation In the dis· course of oral Interviews (as measured by quantity of talk, number of topic Initiations, reactiveness, and topic persls· tence) will vary according to the theme of the Interview.

Many previous studies of IL variation have established that task plays an important role in conditioning variation in IL phonology (Beebe, 1980; Dickerson & Dickerson, 1977; Gatbonton, 1978; Sato, 1985; Schmidt, 1977; Stolen, 1987;

16

I

~

I

I

i

I

j

l l

'• II


II

'• '• ' ' ' ' ' ' ' '

Weinberger, 1987; Wenk, 1986). A smaller number of studies have established a role for task in accounting for variation in IL morphology (Ellis, 1987; LarsenFreeman, 1975; Tarone, 1985; Tarone & Parrish, 1988) and in the syntax ofIL negation (Adamson& Kovac, 1981). HypothesisH6 was formufatedin order to test whether task-related variation in discourse existed across the three tasks in the FCE oral interview. H6.

Dominance, contingency, and goal orientation in the discourse of oral interviews (as measured by quantity of talk, number of topic initiations, reactiveness, and topic persistence) will vary according to the task In which examiner and candidate are engaged. '

Finally, one aspect of the interlocutor effect which may exert an influence on IL discourse is gender. The combinations of males and females in the interview dyads were examined to see ifthere was evidence to support previous findings on the differences in theme-related dominance in NS-NNS interactions between males and between females. This question was formulated as hypothesis H7. H7.

Dominance, contingency, and goal orientation In the discourse of oral interviews (as measured by quantity of talk, number of topic Initiations, reactiveness, and topic persistence) will vary according to the gender of examiner and candidate.

The analytical procedures for testing these seven hypotheses and the results are discussed in the following section.

17

i

.'

I I I

I

Analysis and Results

Simple univariate statistical tests on the distribution of the discourse variables showed that the data are not normally distributed. Standard deviations are high in all cases and the range between maximum and minimum values is also large. More systematic tests revealed that the distributions of most variables are positively skewed and sharply peaked. In addition, the data do not represent a random sample of the population. On the basis of these indications, means and standard deviations are inappropriate in reporting the results. Instead, medians are more appropriate measures of central tendency and nonparametric procedures based on ranks rather than raw scores are used to investigate relations among variables, as recommended by Hatch and Lazaraton ( 1991 ). Table 1 shows the results of four Wilcoxon Rank Sums tests performed to test hypotheses Hl-3 regarding differences between examiner and candidate discourse. As Table 1 shows, the median quantity of talk by candidates in any one . task was 21 t-units, almost twice the median amount of talk by examiners (12.5 t-units). This difference is as predicted by hypothesis Hl and is highly significant (p= .0001). However, the median number of topics initiated by the examiner and by the candidate did not differ significantly. The median number of new topics initiated by the examiner was 5 per task while the median of candidate topic initiations was 7. Thus hypothesis H2, which predicted that examiners will control the discourse through more topic initiations, is not supported. Dominance by the examiner is noticeable, however, in the degree of reactiveness of candidate and examiner to each other's topic initiations. Examiners ratified a median of only 33% of topics initiated by the candidate, while candidates' ratified twice as many (67%) topics initiated by the examiner. This is thedifferencepredicted by hypothesisH3(a) an dis highly significant(p= .0001 ). With regard to the degree to which either party is oriented to internal goals, which we have measured by the persistence of topics they initiate, results show a small but nonetheless significant difference. The median persistence of topics initiated by candidates was 4 t-units, while the median persistence of topics initiated by examiners was 4.5 t-units. This is the difference predicted by hypothesis H3(b). Taken together, these two results show that FCE oral interview discourse is indeed characterized by asymmetrical contingency. However the asymmetry lies primarily in the degree to which the examiner and candidate react to topics initiated by each other and, to a lesser extent, in the degree to which the examiner is more goal-oriented than the candidate. Table 2 shows the median scores for the eight discourse variables related to the examiner in the interview together with results ofKruskal-Wallis tests performed on each variable in order to investigate the relations between examiner and discourse predicted by hypothesis H4. Results show that the examiner does not significantly influence the amount of talk, the number of topics initiated by either

I j.

l

I I

'1·.:

I

•

I'

I I

18

'1

I I

I

1

j

l

'•

ANALYSIS AND RESULTS

II

•II

Table I. Differences between examiner and candidate discourse in 30 FCE oral interviews

Quantity of Talk by

• • •

' 'II

Ill II

'.. II

Percentage of Persistence Topics Topic of Topics Initiated Initiations Initiated by by by Ex. Cand. Ex. Cand. which are Ex. Cand. Ratified

Ex.

Cand.

EQ

CQ

ET

CT

ETR

CTR

ETP

CTP

N

90

90

90

90

90

87

90

87

Median

12.5

21

5

7

67%

33%

4.5

4

Mean Rank

72.7

108.3

85.7

95.3

120.1

56

98.6

79.1

Wilcoxon Z Prob>

IZI

ExaminerCandidate Differences

-4.57

-1.25

-8.23

-2.52

.0001

.2133

.0001

.0116

EQCTR

ETP > CTP

Note. N is the number of tasks. Each interview contained three tasks, for a total of90 observations. However in the case of three tasks the candidate did not initiate any topics, making the values of CTR and CTP meaningless. Thus, only 87 observations are reported for these variables. Significance level set at tt=.05; df=I.

party, nor the persistence of topics in the discourse. Examiners do not vary in their reactivity to candidates' topic initiations; however, as the ETR column of Table 2 shows, candidates do vary significantly in the proportion of topics they ratify from different examiners. Further analysis of the ETR variable by means of Ryan's procedure (Linton & Gallo, 1975) was carried out in order to identify which examiners caused greater reactiveness among candidates7 • Ryan's procedure locates the significant differences among specific pairs of medians when a Kruskal-Wallis test has shown a variable with three or more levels to contribute significantly to variation in the dependent variable. In the case ofETR related to examiner, Ryan's procedure did not find any significant differences among the reactiveness of candidates with different individual examiners .

19

i1


•

Table 2. Median scores and Kru:skal-Wallls tests for interview

Examiner N

EQ

CQ

ET

CT

ETR

CTR

ETP

CTP

A

12 15.5

21

7

5.5

75%

50%

4.8

4.6

. !,

B

21

9

21

4

7

71%

22%

3.5

4

11

c

9

13

21

4

5

50%

30%

4

3.8

11

D

6 16.5

17

5

8

45%

20%

3.9

3.8

E

12

18

26.5

8

8

77%

36%

5.7

3.5

!

F

15

12

16

4

5

75%

33%

6

4

I

G

6

14

23.5

6.5

8

59%

40%

3.8

2.3

H

9

10

19

6

4

83%

25%

5.3,

4.7

discourse related to examiner

I

I

rl

I

ii :I 'I

'II

Kruskal-Wa:Jis 9.90 4.25 Chi square

I

p Eta squared test of association

n.s.

n.s.

7.67

4.46 21.72 13.25 13.40

6.29

n.s.

n.s.

.0028

n.s.

n.s.

n.s.

.24

In order to evaluate the strength of the association between examiner and topic ratifications by the candidate, an 1']2 test of association was performed (Hatch & Lazaraton, 1991). This test showed that only 24% of the variation in ETR can be accounted for by the examiner. There are thus clearly other, as yet unidentified, factors at work in influencing the reactiveness of candidates to examiners' topic initiations and we would be wise not to overestimate the role played by the examiner. Table 3 and Figure 3 show results of the investigation of the re)ationship between interview discourse and the theme of the interview predicted by hypothesis H6. The investigation of hypotheses H5-7 was carried out in a similar way to that reported for hypothesis H4. Tables 3, 4, and 5 report median values · of the discourse variables and the results ofKruskal-Wallis tests in a similar way to Table 2, while Figures 3, 4, and 5 report the results of Ryan's procedures performed in order to locate the differences among medians for those variables on which contextual factors had a significant effect.

,1'I I

11

I:t :

·:I,. i

I I

:I

20

I t

I

I

I

i

Note. For Tables 2-5, significance level set at a~.05; df~7. Kruskal-Wallis statistics were computed on Wilcoxon Scores (rank sums). Median scores are shown in Tables 2-5 rather than rank sums because they are more informative.

·!

~

1

~

I I

~

i

I l I

I


II

• •II

Table J, Median scores and

Theme

N

EQ

CQ

ET

CT

ETR CTR

ETP

CTP

Learning

36

14.5

21

5

6.5

66% 33%

4

4

Having a Good Time

21

11

18

6

8

75% 25%

4.5

3.6

Work

33

12

21

5

6

71% 33%

6

4.4

Kruskal-Wallis Chi square

.36

.65

1.14 3.69 2.04

.53

9.25

.50

p

n.s.

n.s.

n.s.

n.s.

.0098

n.s.

Kruskal-Wallls tests for Interview discourse related to theme

I

•

n.s.

n.s.

Eta squared test of association

• • '• ' II II II

' .10

Table 3 shows that the theme of the interview, as determined by the choice of the interaction pack, had a significant effect on only one discourse variable-the persistence of topics initiated by the examiner. Figure 3 shows that the theme of "Work" produced significantly longer topics than the theme of "Learning" but that the length of examiner-initiated topics in interviews on thetheme of"Having a Good Time" did not differ significantly from interviews on either of the other two themes. The median persistence of examiner-initiated topics in interviews on the theme of "Work" was 6 t-units, while the median persistence of examinerinitiated topics in interviews on the theme of "Having a Good Time" was 4.5 tFigure J, Differences among median ETP scores related to theme obtained by Ryan's procedure. Medians within the same box are not significantly different at a=

.OS.

ETP

Higher

Work Having a Good Time

Lower

'

Learning

21


Table 4. Median scores and

Task

N

EQ

CQ

ET

CT

ETR

CTR

ETP

CTP

Photographs Task

30

12

21

5

7

77%

40%

4.7

5

Passages Task

30

7

10

3

2.5

67%

20%

4.1

2.5

30

20.5

35.5

8

10.5

67%

32%

5

4


44.12 50.68 31.48 37.70

4.34

5.62

.73

18.51

p

.0001 .0001 .0001 .0001

11.S.

11.S.

11.S.

.0001

Kruskal-Wallis tests for Interview discourse related to task

Ranking

Activities

Eta squared test of association

I

.50

.57

.35

.42

.21

units and "Learning" was 4 !-units. However, an l)2 valueof10% shows that there is a weak association between theme and ETP and that other factors must also be influencing topic persistence. The influence of task on the discourse is shown in Table4 and Figure 4. Task appears to have the strongest effect of any of the contextual variables considered but its effect is most apparent in the measures of dominance. The quantity of talk from both examiner and candidate, the number of topics initiated by both parties, and the persistence of topics initiated by the candidate are all affected by the task in which examiner and candidate are engaged, as predicted by hypothesis H6. However, no significant relationship exists between task and the two measures of contingency (ETR and CTR), nor the persistence of topics initiated by the examiner. The results of Ryan's procedure reported in Figure 4 show that on the measures of quantity and topic initiations, all three tasks differed significantly from each other, with the ranking activities producing the greatest quantity of talk and the greatest number of topic initiations; the photographs task producing the least; and the passages task producing an amount intermediate between the other two. As far as the persistence oftopics initiated by the candidate is concerned, the photographs task produced significantly longer topics (a median of 5 t-units) than the other two. There was no significant difference in the persistence of candidateinitiated topics produced by the ranking activities or the passages task. Tests of association by means of the l)2 statistic between task and the five discourse variables on which it exerted a significant effect show a moderate to strong association withEQ ( 50%) and CQ (57%), a weaker association with ET

1'

''

22

~

I

I '

I

t I t

I


II

II Ill II

• •II • •Ill • • •

Figure 4. Differences among

median EQ, CQ, ET, er, and CTP scores related to task obtained by Ryan's procedure. Medians within the same box are not signlficantly different at ex= .05.

Higher

I I I I I I I I I

Lower

EQ

CQ

ET

CT

CTP

Ranking Activities

Ranking Activities

Ranking Activities

Ranking Activities

Photographs Task

Photographs Task

Photographs Task

Photographs Task

Photographs Task

Ranking Activities

Passages Task

Passages Task

Passages Task

Passages Task

Passages Task

(35%) and CT(42%), and a weak association with CTP (21%). It should be recalled at this point that in the format of the interview, the three tasks always occuTI"ed in the same order within one interview, namely, first photographs, second passages, and last ranking activities. It is thus impossible to separate the effect of task from a serial effect. It may well be that by the third task, the candidate and examiner are more relaxed than they were in the early stages of the interview and are thus able to speak more freely and initiate more topics than in the previous tasks. Without fmiher studies in which the tasks are presented in random order, it is not possible to say whether the manifestly different discourse produced on the three tasks is due to the nature of the tasks themselves or to a serial effect or, as is most likely, to a combination of the two. Finally, we tum to the effect of gender on the interview discouFse predicted by hypothesis H7. Table 5 shows that the gender of examiner and candidate significantly affect two aspects of the discourse-the persistence of examinerinitiated topics and theirratification by the candidate. Gender did not significantly affect quantity or number of topic initiations from either party, neither did it affect the persistence of candidate-initiated topics nor their ratification by the examiner. As Figure 5 shows, interviews between women produce significantly more reactiveness by candidates (82%) than interviews between male examiners and female candidates (64%). Interviews between female examiners and male candidates and those between two men did not differ significantly from the other gender combinations on this variable. As regards the persistence of examinerinitiated topics, again the interviews between women produced significantly longer topics (a median of 6.2 t-units) than interviews between a male examiner and a female candidate (4 t-units) or between two men (3.7 t-units). The 23

I I I I


Table 5. KruskalMWallls tests for

Gender

Interview discourse related to gender

Cand

N

EQ

CQ

ET

CT

ETR

CTR

ETP

CTP

F

21

16

19

6

5

82%

33%

6.2

5.5

M

15

12

21

5

7

67%

29%

4.8

3

F

33

13

22

5

8

64%

33%

4

4

M

21

11

19

4

6

67%

24%

3.5

3.7


3.40

2.29

2.76

1.99

13.57

4.23

16.79

3.40

p

n.s.

11.S.

11.S.

11.S.

.0036

n.s.

.0008

11.S.

Ex

F

M

Eta squared test of

.15

association

.19

persistence of topics initiated by a female examiner in interviews with a male candidate did not differ significantly from that produced by the other gender pairings. However, r\2 tests of association reveal that only 15% of the variance in ETR and 19% of the variance in ETP is attributable to gender. Other factors in addition to gender are clearly at work here. The significance of these results will be discussed in the remaining sections. Figure 5. Differences among median ETR and ETP scores

ETR

ETP

FF

FF

l

FM

FM

1

MM

MF

MF

MM

:I

i

I

related to gender of examiner and candidate obtained by Ryan's procedure. Medians within the

same box are not significantly

different at a= .OS.

Higher

i I

Lower j.i

'

I

24

'I

I I I I

1

I I

t

I

I

Discussion

• • • • •II • •I •I '

•

These results highlight several interesting features ofNS-NNS interactions in oral proficiency interviews. First of all, the analysis confirms what scholars had predicted about interview discourse. The NS examiner and NNS candidate make radically different contributions to the discourse. The degree to which the examiner exercises control over the discourse is apparent from the much greater reactiveness of candidates to examiners' topic initiations than vice versa, and from the slightly longer persistence of examiner-initiated topics. The discourse we · observe in theseinterviews shows evidence ofgreater goal orientation by one party and much greater reactiveness by the other party. It thus clearly exemplifies the style that Jones and Gerard have called asymmetrical contingency. However, when we look at the measures of dominance-quantity of talk and number of topic initiations-the data appear to suggest that the candidate dominates the interview since the candidate talks about twice as much as the examiner, while there is no significant difference between the number of topics initiated by either party. These measures of dominance do not capture the underlying control over the right to speak since apparently the candidate has the right to speak more than the examiner. However, the results for reactiveness and persistence show that, nonetheless, this right is limited by the examiner. According to the rules of the interview, the examiner gives the floor to the candidate most of the time but still controls what is spoken about and is able to take or withhold the floor at any time. The greater reactiveness of the candidate is evidence that this control over the discourse is accepted. Jones and Gerard and van Lier claim that "conversation" involves more balanced reactiveness by both parties to each other's topic initiations. However, since this is the first study to operationalize this model, we still lack an analysis of baseline NS-NNS conversation data which would allow us to make a true comparison. A particular strength of the contingency model has been in demonstrating that dominance in these interviews is not simply reflected by the relative amount of talk by participants nor by the relative numbers of topic initiations, as several earlier studies of dominance had claimed. Instead, dominance can be seen in these interviews in the degree ofreactiveness ofthe candidate to the examiner's initiations, but not vice versa . Ifwe tum our attention now to the contextual factors which affect the structure of the discourse, we see that different contextual factors affect participants differently and this may be accounted for by participants' different roles in the interview. First, those contextual factors which affect only candidate discourse will be discussed; second; those contextual factors which affect only examiner discourse will be discussed; finally, those contextual factors affecting both candidate and examiner equally will be considered. 25

DISCUSSION

Contextual Influences on the Candidate The reactiveness of the NNS candidate to topics initiated by the NS examiner appears from Tables 2 and 5 to be related to certain individual differences among examiners. Female candidates are more reactive to topics initiated by female examiners and all candidates appear to vary in the degree to which they ratify topics initiated by different examiners. There is probably some interaction between ETR due to examiner and ETR due to gender here since examiners H, E, and F, who induced the highest ETR from candidates, are aU female, while examiners A, B, G, C, and Dare male and they induced the lowestETR. However, other individual differences between examiners may be relevant here such as age, empathy, and other characteristics which lead to greater solidarity between examiner and candidate. These characteristics were not measured in this study The persistence of topics initiated by the NNS candidate appears to be due to task. The photographs task and theranking activities produce candidate-initiated topics which persist about twice as long as topics initiated by the candipate in the passages task. What makes the passages task so different from the other two? This task involves the candidate matching a written text with a photograph and is unique in certain respects since it does not aUow for a negotiated outcome (Rulon & McCreary, 1986; Young, 1984, 1988). In the passages task there is only one right answer and this is known in advance to the examiner-the correct match of text and photograph. The examiners' topic initiations in the passages task are thus display questions while in the other two tasks in the interview the examiners pose very few display questions. In the other two tasks there are no predetermined outcomes and candidates are thus able to negotiate the outcome with examiners. Previous research on ESL classroom discourse by Brock (1986) has shown that students' responses to teachers' display questions are significantly shorter than responses to referential questions and a similar effect is clearly visible in the present results. Contextual Influences on the Examiner The examiner's contribution to the discourse appears to be affected by contextual factors in ways different from the candidate. The examiner's degree ofreactiveness (CTR) is not affected by any contextual factors. Howev~r,-the goal orientation oftheexaminer evidentinETPappears to be affected by gender as weU as by the theme of the interview. The effect of gender on examiner topic persistence may be related to thereactiveness ofthe candidate, since both ETPand ETR are significantly higher only in interviews between women. Since topic persistence is counted across turns by both participants, it wiU be higher if one

i,\ 1;

,,

II'

I

'i '11!

26

• •I

I I I I I II

I

I I I

I

DISCUSSION

participant ratifies the topics introduced by the other as appears to be the case here. On the other hand, if an examiner-initiated topic is not ratified by the candidate the topic dies and ETP will have a value approaching unity. Thus greater reactiveness by the candidate produces longer topics initiated by the examiner. The theme of "Work" produces significantly longer examiner topics than the theme of "Leaming." This thematically related difference in the discourse may be due to the fact that the candidates in the interviews were mostly young adults, many of whom were looking forward to graduating soon from college or university. A favorite set of topics for discussion in many interviews with these candidates centered on their future plans for a career after graduating. Examiners often initiated topics in this area by asking candidates to talk about their career plans. Such topics persisted in the discourse because candidates had a lot ofideas and opinions about the subject. The emotional investment ofthe candidates in the theme of work at this particular stage in their lives may explain the link between theme and topic persistence. As has been mentioned above, previous studies have shown a relationship between a NNS's expertise on a given theme or emotional investment in the theme and the structure ofNS-NNS interactions. The present study did not measure independently candidates' expertise or emotional investment in the three themes and we are thus only able to speculate on the reasons for the apparent effect of theme on the discourse. We look forward to further studies of oral proficiency interviews which investigate this relationship in greater detail.

Contextual Influences on the Discourse as a Whole Of all the contextual factors that were investigated, task exerts the strongest influence on the structure of the discourse in these interviews. Agreat deal of the variance in the quantity of talk from both parties is due to task. Variance in the number of topic initiations by both parties is also influenced to a significant degree by task, albeit less so than with quantity. Moreover, the influence of each of the three tasks is significantly different one from the other, with the ranking activities producing the highest EQ, CQ, ET, and CT, the passages task producing the lowest, and the photographs task producing an amount intermediate between the other two. We have already discussed the possible reasons why the passages task stands out as producing the shortest topics and the results for quantity and topic initiations may be explained in the same way. The passages task does not allow candidates to negotiate the outcome of the task and results in display questions from the examiner. Both these factors combine to create very short interactions. However, what causes the greater quantity and higher number of topic initiations on the ranking activities in comparison to the photographs task? We believe these results are due both to a serial effect and to the nature of the ranking 27

D1scuss10N

activities task. The ranking activities are the last task in the interview and the photographs task is the first task. As we have suggested above, by the later stages of the interview examiner and candidate may be more relaxed in each other's company and this may lead both parties to speak more freely and at greater length. In addition, the ranking activity is the only one in which candidates are asked to talk about their own beliefs and their own experiences. While this may also happen to a certain degree in discussing thephotographs, that task and the passages task center primarily on the discussion of items which are elicited by the interaction pack materials which candidates have never seen before the interview. In the third activity, therefore, candidates may simply havemore to talk about than in the other two and this is reflected in the higher values for quantity and topic initiations on this task. Finally, it should be noted that task, while exe1ting a strong influence on the quantity of talk and the number of topic initiations by both parties as well as on the persistence of candidate-initiated topics, does not exert a significant effect on the reactiveness of either the candidate or the examiner. This suggests an interesting distinction between personal contextual variables such as gender and other individual differences among examiners on the one hand and task-related contextual variables such as task and theme on the other. Discourse in these interviews is mediated by permanent personal characteristics of examiners and candidates such as gender, and by temporary situational factors such as task and theme. The reactiveness of a candidate in ratifying topics initiated by the examiner depends on who the candidate is speaking to, while the quantity of candidate talk, the number of topics initiated by the candidate and their persistence depends on situational factors not related to the person to whom she is speaking. In other words, reactiveness appears to be interpersonal, while the other features of the discourse appear to be situational in origin.

i

'I

,I

I

I

i

• •I

I I I I I

I I f

i,;

,J :1

'I

'I

28

'

Conclusions

I I

I

This study is principally exploratory and descriptive. There are remarkably few descriptive linguistic stiidies of discourse in oral proficiency interviews. Current work under way by researchers working within a number of different theoretical perspectives (Lazaraton, 1991; Ross & Berwick, 1991; Yoshida, 1991) promises to greatly increase our understanding of the phenomenon. But given our present limited knowledge of oral interview discourse, the present study is also limited in a number of ways. The study is descriptive rather than experimental since a nonrandom sample of existing interviews was studied and we have no knowledge of several ofthepotentiallyrelevantvariables in the study, such as learners' expertise or emotional investment in the theme ofthe interview, examiners' empathy or age, or the potential confound between task and a serial effect. Results from this study should therefore be treated as suggestive rather than conclusive. In addition, although the contingency model of dyadic interaction has very effectively summarized the overall structure of the discourse, it is nonetheless a static rather than a dynamic description. That is to say, it does not reveal the process by which discourse is created through interaction but only gives us a snapshot of the product of that process. We need to go into much greater detail in describing the nature of contingency as it exists between each set of turns in order to understand how discourse is created. The strengths of the study lie in describing the discourse of oral proficiency interviews in te1ms of a general model of dyadic interaction because in so doing we have illuminated some important properties of the discourse. Dominance in these dyadic interactions is not simply a question of speaking more or of initiating more topics but has to do rather with the goal-orientation of the examiner and the reactiveness of the candidate. Dominance is also evident in the asymmetry of the interaction: the examiner is far less reactive than the candidate and the candidate appears a little less goal-oriented than the examiner. We may ass\.Jme, following Jones and Gerard and van Lier, that this kind of discourse is a unique product of the interview situation. Conversations involving NNSs in other situations may be very different in structure from the discourse of oral proficiency interviews. Ifthis is the case, then van Lier's witty characterization of the peculiarities of oral interview discourse as "reeling, writhing, drawling, stretching and fainting in coils" is apt indeed. The oral proficiency interviews that have been examined here bear very little resemblance to what we believe to be the structure of conversation. The interviews often fail to elicit discourse from NNS candidates such as maintaining a topic over a large number of turns that we may want to consider in assessing oral proficiency. At the same time, the interviews often confront candidates with situations such as having to respond to rapidly shifting topics initiated by the NS, which under other circumstances we may consider of little relevance in the assessment of oral proficiency. 29

,

DISCUSSION

We have also seen that there exists significant variation in the discourse. This variation has been sttributed to the effect of the examiner, the task, the theme of the interview, and the gender of the participants. Since task appears to contribute most to the structure of the discourse, it is probably a good idea that the FCE interviews sample candidates' performance over a range of three different tasks. Still, task is not all. There are some interpersonal features of the candidate's discourse such as reactiveness which appear to be unaffected by task but sensitive instead to the learner's interlocutor. Gender appears to play a role in this, but it would be worthwhile to conduct further investigations into examiners' interactional styles to see what other factors cause greater or lesser reactiveness among candidates. Yet, for all this detailed description of interview discourse and the variation in it, we still do not know whether any of this makes a difference to the examiner's assessment of the learner's oral proficiency. Now that we have a working model of interview discourse which both descnbes and explains important aspects of the linguistic interaction, we are in a good position to see whether variations in the discourse affect examiners' judgments. Further research along these lines will help us to understand what native speakers react to when they judge the oral proficiency of a learner of their language and-most interestingly-the nature of oral proficiency itself.

I

i

:

I I I

'' !

l

JO

II

.. .. II II II

Recommendations for Future Research and Development The research which is described in this report is, as has been stressed, deliberately limited in scope. In essence, theprojecthas established a workable and insightful model of dyadic discourse in oral interviews between native and non-native speakers of English. It has been shown that the structure of the discourse as described by this model varies according to task, theme, and interlocutor. In these respects, interlanguage discourse has shown itself to be subjectto similar contextual influences as interlanguagephonology, morphology, and syntax. Although the results described here are limited by the pilot nature of the study, they are also suggestive of further applied and theoretical research into the nature of oral proficiency in a second language. In this final section, possible directions for future research will be sketched in five areas: (a) further refinements of the model of discourse developed here; (b) extension of the model to comparative descriptions of other genres of oral interaction in addition to interviews; (c) further investigations of the reasons underlying the effects of task, theme, and interlocutor on the discourse; (d) extension of the methods and models of this study to investigate oral interviews including learners from other first language backgrounds and following different interview formats; and (e) finally, the development of new methods and the adaptation of existing techniques of assessing oral proficiency in the context of a large scale testing procedure such as the First Certificate in English.

Further Development of the Model The model ofdyadic discourse developed here rests on a clear and unambiguous definition of topic. However, as has been mentioned earlier in this report, for many reasons the notion of topic is by no means an easy one to work with. In order to develop the model further, therefore, a better definition of topic is needed and explicit procedures for identifying topics in spoken discourse should be worked out. The analysis of discourse presented in this report is static rather than dynamic-a retrospective summary of discourse achieved rather than an ongoing description ofhow that discourse develops. For many reasons, a more developmental approach to discourse analysis is desirable which will show, for example, how patterns ofcontingency, goal-orientation, and dominance change within one discourse task. The model, as it stands, has been developed and applied to dyadic discourse, yet the principles of contingency, goal-orientation, and dominance are equally applicable to interactions involving more than two participants. Further

31

I RECOMMENDATIONS FOR fUTURE RESEARCH AND DEVELOPMENT

development of the model to handle multi-party interaction may be particularly useful in order to compare multi-party FCEoral interviewswith the dyadic data analyzed here.

Comparative Analysis of Other Discourse Genres 'I

Much of the argument put forward regarding the weaknesses of the oral interview procedure in the present report rests on a comparison between the discourse of oral interviews and other genres of oral interaction, including informal conversations between native and non-native speakers. However, while a thorough description of interview discourse now exists, baseline data are lacking on other genres of oral interaction using the same model of description. An important priority is thus to conduct further research using this model into interactions between native and non-native speakers of English in other situations such as classrooms, informal social activities, and service encounters. This needed research will provide a solid foundation for establishing the validity of oral interview procedures.

'

I

!I . I

I .

I' ,

'

I I I

'I

'

Further Investigation of Task, Theme, and Interlocutor

, I

Since the present investigation is descriptive in nature, the underlying causes of the observed effects of task, theme, and interlocutor on the structure of the discourse can only be surmised. The effect of task is confounded in the present study with a serial effect. Further investigation is necessary in order to isolate the effect of task. The theme of the interviews has been shown to affect the discourse and it has been suggested that candidates' expertise and emotional investment in different themes is the cause for this. However, only further studies in which these variables are controlled will allow us to determine the underlying cause of theme-related variation in discourse. The identity of the examiner and the gender of examiner and candidate also affect the discourse but we have no information in this study regarding possible reasons why this should be the case. Speech accommodation theory suggests that when participants share social identities including such factors as age, social class, gender, and ethnicity, the discourse they create may reflect this shared identity and differ significantly from discourse in which such factors are not shared by participants. In order to investigate whether similar or different social identities underlie the observed interlocutor effect in this study, much more information is needed about both examiners and candidates than was available in the present study.

I

,I I

I

32

I

RECOMMENDATIONS FOR FUTURE RESEARCH AND DEVELOPMENT

The investigation of the causes of variation can only be carried out through studies involving careful control of the underlying variables. The variables identified in the present study of task, theme, and interlocutor may be seen as mediating (and masking) the underlying roles of these variables.

Comparative Analysis of Other Interview Formats

II

• •II I

~ ~

The present study has been limited to interviews involving candidates from only two first language backgrounds-Italian and Brazilian Portuguese-and examiners who had considerable experience of living in those countries. The familiarity of the cultures of these candidates to the examiners may produce greater understanding between the participants than would exist between British examiners in the U.K. and candidates from non-Western countries such as Japan or the Arab world. On the other hand, the format of the interview may be so familiar to all well-prepared candidates and to trained examiners that no culture-related differences of understanding exist. The question of whether cultural differences in interactional style affect the discourse of oral proficiency interviews is an open one, but one which cannot be ignored in an interview which is conducted world wide. The conclusions of the present report are based on interviews from a deliberately limited and selected sample of interviews involving candidates from only two countries and expatriate examiners. Further research involving random sampling of interviews world-wide is necessary in order to generalize these findings to all interviews and all candidates. The Cambridge FCE oral interview is a highly structured format involving three separate tasks and contrasts significantly with other major international oral proficiency interviews such as the Language Proficiency Interview (LPI) originally developed by the U.S. Foreign Service Institute. In the LPI (and its domestic cousin, the oral interview of the American Council of Teachers of Foreign Languages), the whole interview follows an extended question-answer format without stimulus materials or the varied tasks of the FCE oral interview. A comparative analysis of the discourse from both kinds ofinterviews-FCE and LPI-would be instructive in suggesting ways in stimulus materials and particular tasks stimulate or inhibit conversation-like discourse.

33


Oral Interview Design Considerations Finally, the issue of greatest practical significance for the Syndicate is how the knowledge of oral interview discourse derived from the present study may be utilized in the design of new methods of oral assessment in the future or in the adaptation of existing techniques in order to maximize reliability and validity of the assessment procedure. Given that the present study has identified task as the single most important factor in shaping interview discourse, top priority should go to research into the design oftasks which elicit the kind of discourse which is desired. In the present FCE oral interview format, the Passages Task stands out from the other two as the shortest task and the one in which the examiner dominates the discourse the most. It may be worthwhile to consider ways of replacing or enhancing this task to ensure a more balanced interaction. The influence of the examiner on the discourse of oral interviews has been established in the present study, in paiticular on the reactivity of the candidate. This suggests that examiners could perhaps be trained to stimulate greater reactivity on the part of the candidate and thus to create a more balanced discourse. This study did not investigate the relationship between candidates' proficiency in English and the interview discourse, but recent research on noviceexpert differences in proficiency (Mislevy, 1989; Snow, 1990; Yoshida, 1991; Young & Perkins, 1991) suggests that increasing proficiency will not necessarily imply a concomitant increase in candidate discourse variables. Rather, the configuration of factors underlying oral proficiency for beginners may differin complex ways from the configuration of factors for advanced learners. Further research into novice-expert differences in oral second. language proficiency has important implications for the design of rating scales used in oral proficiency interviews. The current rating scales oftheFCE oral interview focus on fluency, grammatical accuracy, pronunciation of sentences and individual sounds, interactive communication, and vocabulary. It may be desirable to redesign these scales in order to take into account our greater understanding of the structure ofNS-NNS interview discourse. Finally, we know very little of how native-speaking examiners' judgments of oral proficiency are influenced by non-native-speaking candidates' interactional style. Since examiners are involved with candidates in the construction of discourse in the FCE oral interview, their initial judgments of candidates' proficiency as reflected in candidates' interactional style may lead examiners to modify their own interactional style and thus restrict the candidate's opportunity to interact in a style appropriate to a different level of proficiency 34

• • • • •I •I

I

'I I I

I I I I I I I I I I I I


from the one at which they were initially judged to be. This self-fulfilling prophecy may contribute to less reliable judgements of proficiency by the examiner. One way to improve such a situation is to make examiners conscious of the role of the candidates' interactional style in influencing their judgments. The present study has shed some light on the nature of oral proficiency in a second language and on the practical endeavor of assessing it. However, as reviewed here, there are many further avenues ofinvestigation to pursue before we have a full understanding of oral proficiency and before we can devise means of assessment appropriate to that understanding.

l

35

Notes I. Parts ofthis report were presented atthe 1991 Second Language Research Fonun, held February 28-March 3 at the University of Southern California, Los Angeles and at the 25th Annual TESOL Convention, held in New York, March24-28, 1991. The help of Michael Milanovic in conceptualizing this project and in guiding it to fruition is gratefully acknowledged. The following scholars also provided valuable ideas and criticisms: Evelyn Hatch, Joyce Neu, Kyle Perkins, and Jane Zuengler.

2. For the sake of terminological clarity it is necessary to distinguish between two sensesofthetenn "topic." Unfortunately, conventionaldefinitiousoftopicas "what is talked about" (Richards, Platt, & Weber, 1985) do not specify the stretch of utterance to which a topic applies and this has Jed to considerable terminological confusion over what is already a difficultterm to define and use. In this paper the term "topic" will continue to refer to what is talked about in one t-unit or clause. However, the term "theme" will be used to refer to what is talked about in a longer stretch of discourse such as one of the interviews. Thus we will refer to this picture as the topic of the utterance "Tell me about this picture," while we will refer to "Learning," "Having a Good Time," or "Work" as the theme ofthe interviews discussed below. Topic, as used in this paper, thus corresponds loosely to the sentential topic described by Brown and Yule (1983), while theme corresponds loosely to what Brown and Yule call the title of a stretch of discourse.

• •I • •I • •I

3. Jones and Gerard use the term "planning" to refer to the degree to which each participant in an interaction attempts to achieve bis orber own goals. Planning is a term which has become widely used in second language acquisition in the sense in which it was used by Ochs (1979) to refer to forethought and organizational preparation in advance of speech. While planning in Ochs' sense surely coincides with the goal-orientation of the Jones and Gerard model, goal-orientation in interaction does not necessarily imply forethought and advance preparation. Thus, to avoid possible confusion, "goal orientation" will be used in this paper to refer to what Jones and Gerard have called "planning." 4. All illustrative quotations are taken from the Cambridge First Certificate in English oral interviews collected as part of this study. Participants in the interviews are the native-speaking examiner, and the non-native-speaking candidate. Transcription conventions are as follows. A short pause or tone group bow1dary is indicated by " . " and a long pause is indicated by " .. " including cases when such pauses occur between turns. When participants are speaking at the same time, the beginning and end of theiroverlapping speech is indicated by vertical lines in the transcript. When a speaker continues to bold the floor during an interruption, comment, or backchannel cne from another speaker, the continuation ofthe first speaker's turn is indicated by"=" and by beginning the firstwordoftbespeaker's continued turn with a lower case letter. Unclear or Wlintelligible speech is indicated by "xxx". Heavy stress, often used in contrasts, is indicated by underli!!ing the syllable stressed. The only other punctuation used is a question mark to indicate a rising intonation and a capital letter at the beginning of each new tum. 5. Haviland and Clark (1974) have proposed a Given-New Strategy for comprehension which also explains some of the difficulty that learners experience in comprehending utterances. According to Haviland and Clark, in comprehending an utterance listeners search memory for antecedent infonnation (from the immediately preceding discourse) and then attempt to relate the infonnation in the new utterance to the given infonnation in the preceding 36

II

• ' ' ' '

NOTES

utterance. In cases of disjunctive contingency between utterances, the connection ofnew to given information is hard to make and the listener, especially a NNS, may fail to comprehend the new utterance. 6. See note 2 for a discussion of the scope of topic as a discourse writ and the distinction made here between topic and theme. 7. Ryan's procedure isa nonparametric test for locating differences among medians. It is analogous to the Scheffel multiple-range test for means found to be significant by

parametric analysis of variance. See Hatch and Lazaraton (1991, pp. 337-340).

37

w·"-.,- --- - -

References Adamson, H. D., & Kovac, C. (1981). Variation theory and second language acquisition: An analysis of Schumann's data. In D. Sankoff & H. J. Cedergren (Eds.), Variation omnibus (pp. 285-292). Edmonton, Alberta: Linguistic Research. Argyle, M., & Dean, J. (1965). Eye contact, distance and affiliation. Sociometry, 28, 289-304. Beebe, L. M. (1977). The influence of the listener on code-switching. Language Learning, 27, 331-339. Beebe, L. M. (1980). Sociolinguistic variation and style shifting in second language acquisition. Language Learning, 30, 433-448. Beebe, L. M., &Zuengler, J. (1983). Accommodation theory: An explanation for style shifting in second language dialects. In N. Wolfson & E. Judd (Eds.), Sociolinguistics and language acquisition (pp. 195-213). Rowley, MA: Newbury House. Berkowitz, D. (1989). The effect of cultural empathy on second-language phonological production. In M. R. Eisenstein (Ed.), The dynamic interlanguage: Empirical studies in second language variation (pp. 101-114). New York: Plenum. Brock, C. A. (1986). The effects of referential questions on ESL classroom discourse. TESOL Quarterly, 20, 47-59. Brown, G., & Yule, G. (1983). Discourse analysis. New York: Cambridge University Press. Clarke, D. D., & Argyle, M. (1982). Conversation sequences. In C. Fraser & K. R. Scherer (Eds.), Advances in the social psychology of language (pp. 159-204). New York: Cambridge University Press. Dickerson, L. J., & Dickerson, W. B. (1977). Interlanguage phonology: Current research and future directions. In S. P. Corder & E. Roulet (Eds.), Actes du 5eme Colloque de Linguistique Appliquee de Neucha-

I 11

I I' 111

I! j, qi I I ,I I

i!i

i

'

'•II"

[!

tel: The notions ofsimplification, interlanguages andpidgins and their relation to second language learning (pp. 18-29). Geneva: Droz. Eisenstein, M. R., & Starbuck, R. J. ( 1989). The effect ofemotional investment

'I 1,

11

I

i 11

on L2 production. In S. Gass, C. Madden, D. Preston, & L. Selinker (Eds.), Variation in second language acquisition volume II: Psycholinguistic issues (pp. 125-137). Clevedon, Avon: Multilingual Matters. Ellis, R. (1987). Interlanguage variability in narrative discourse: Style shifting in the use of the past tense. Studies in Second Language Acquisition, 9, 1-20. Erickson, F., & Shultz, J. (1982). The counselor as gatekeeper: Social interaction in interviews. New York: Academic Press.

I I !

II \ I'' ,'' I ,,

Iii,' i.'

r

I'

Ii

38

I I l

I I

II I I I I I I I

II

REFERENCES

Gaies, S. J. (1982). Native speaker-nonnative speaker interaction among academic peers. Studies in Second Language Acquisition, 5, 74-81. Gass, S. M., & Varonis, E. M. (1986). Sex differences in nonnative speakernonnative speaker interactions. In R. R. Day (Ed.), Talking to learn: Conversation in second language acquisition (pp. 327-351 ). Rowley, MA: Newbury House. Gatbonton, E. (1978). Patterned phonetic variability in second language speech: A gradual diffusion model. Canadian Modern Language Review, 34, 335-347. Giv6n, T. (1984). Universals of discourse structure and second language acquisition. In W. E. Rutherford (Ed.), Language universals and second language acquisition (pp. 109-136). Philadelphia: John Benjamins. Gumperz, J. J. ( 1979). The retrieval of sociocultural knowledge in conversation. Poetics Today, 1, 273-286. Haviland, S. E., & Clark, H. H. (1974). What's new? Acquiring new information as a process in comprehension. Journal of Verbal Learning and Verbal Behavior, 13, 512-521. Hatch, E., & Lazaraton, A. ( 1991 ). The research manual: Design and statistics for applied linguistics. New York: Newbury House. Jones, E. E., & Gerard, H.B. (1967). Foundations ofsocialpsychology. New York: Wiley. Lazaraton, A. (1991). Interaction in the language interview: A conversation analytic perspective. Unpublished doctoral dissertation, University of California at Los Angeles. Larsen-Freeman, D. (1975). The acquisition of grammatical morphemes by adult ESL students. TESOL Quarterly, 9, 409-419. ' Leet-Pellegrini, H. M. (1980). Conversational dominance as a function of gender and expertise. In H. Giles, W. P. Robinson, & P. M. Smith (Eds.), Language: Social psychological perspectives (pp. 97-104). New York: Pergamon. Levinson, K. ( 1991 ). Cooperation and dominance in second language learner interactions. Paper presented at the 1991 Second Language Research Forum, University of Southern California, Los Angeles, February 28-March 3. Linton, M., & Gallo, P. S. (1975). The practical statistician: A simplified handbook ofstatistics. Monterey, CA: Brooks/Cole. Long, M. H. (1983). Native speaker/non-native speaker conversation and the negotiation ofcomprehensible input. Applied Linguistics, 4, 126-141. 39

REFERENCES

Mislevy, R. J. (1989). Foundations of a new test theory. Princeton, NJ: Educational Testing Service. Neu, J. (1986). American English business negotiations: Training for nonnative speakers. English for Specific Purposes, 5, 41-57. Ochs, E. (1979). Planned and unplanned discourse. In T. Giv6n (Ed.), .syntax and semantics. Volume 12: Discourse and semantics(pp. 51 ~80). New York: Academic Press. Richards, J., Platt, J., & Weber, H. (1985). Longman dictionary of applied linguistics. Harlow, Essex: Longman. Ross, S., & Berwick, R. 1991. The discourse of accommodation in oral proficiency examinations. Manuscript submitted for publication. Rulon, K. A., & McCreary, J. ( 1986). Negotiation of content: Teacher-fronted and small-group interaction. In R. R. Day (Ed.), Talking to learn: Conversation in second language acquisition (pp. 182-199). Rowley, MA: Newbury House. Sato, C. J. (1985). Task variation in interlanguage phonology. In S. M. Gass & C. G. Madden (Eds.), Input in second language acquisition (pp. 181-196). Rowley, MA: Newbury House. Schegloff, E. A. (1978). On some questions and ambiguities in conversation. In W. U. Dressler (Ed.), Current trends in textlinguistics (pp. 81-102). New York: Walter de Gruyter. Schmidt, R. W. (1977). Sociolinguistic variation and language transfer in phonology. Working Papers in Bilingualism, 12, 79-95. Schneider, M., & Connor, U. (1990). Analyzing topical structure: Not all topics are equal. Studies in Second LanguageAcquisition, 12 ,'411-427. Selinker, L., & Douglas, D. (1985). Wrestling with 'context' in interlanguage theory. Applied Linguistics, 6, 190-204. Selinker, L., & Douglas, D. (1989). Research methodology in contextuallybased second language research. Second Language Research, 5, 93-126. Smith, J. (1989). Topic and variation in ITA oral proficiency: SPEAK and field-specific tests. In R. Young (Ed.), The training of international teaching assistants [Special issue]. English for Specific Purposes, 8, 155-167. Snow, R. E. (1990). New approaches to cognitive and conative assessment in education. International Journal ofEducational Research, 14, 455-4 73. St0len, M. ( 1987). The effect ofaffect on interlanguage phonology. In G. Ioup & S. H. Weinberger (Eds.), Interlanguage phonology: The acquisition ofa second language system (pp. 389-400). Cambridge, MA: Newbury House. 40

• -I • -.. • • • • -

'

I

REFERENCES

Takahashi, T. (1989). The influence of the listener on L2 speech. In S. Gass, C. Madden, D. Preston, & L. Selinker (Eds.), Variation in second language acquisition. Volume I: Discourse and pragmatics (pp. 245-279). Clevedon, Avon: Multilingual Matters. Tarone, E. E. (1985). Variability in interlanguage use: A study of styleshifting in morphology and syntax. Language Learning, 35, 373-404. Tarone, E. E. (1988). Variation in interlanguage. London: Edward Arnold. Tarone, E. E., & Parrish, B. (1988). Task-related variation in interlanguage: The case of articles. Language Learning, 38, 21-44. University of Cambridge Local Examinations Syndicate. (1982). Cambridge examinations in English: Changes of syllabus in 1984. Cambridge: Author. University of Cambridge Local Examinations Syndicate. (1990). Examinations in English as aforeign language: Instructions to oral examiners. Cambridge: Author. van Lier, L. (1989). Reeling, writhing, drawling, stretching and fainting in coils: Oral proficiency interviews as conversation. TESOL Quarterly, 23, 489-508. Weinberger, S. H. (1987). The influence ofcontext on syllable 'simplification. In G. Ioup & S. H. Weinberger (Eds.),Interlanguage phonology: The acquisition ofa second language system (pp. 401-417). Cambridge, MA: Newbury House. Wenk, B. (1986). Crosslinguistic influence in second language phonology: Speech rhythms. In E. Kellerman & M. Sharwood Smith (Eds.), Crosslinguistic influence in second language acquisition (pp. 120-133). New York: Pergamon. Woken, M. D., & Swales, J. (1989). Expertise and authority in native-nonnative conversations: The need for a variable account. In S. Gass, C. Madden, D. Preston, & L. Selinker (Eds.), Variation in second language acquisition. Volume I: Discourse and pragmatics (pp. 211-227). Clevedon, Avon: Multilingual Matters. Yoshida, Y. (1991). The use of communication strategies use by adult Japanese speakers of English as a second language in language proficiency interviews. Unpublished master's thesis, Southern Illinois University at Carbondale. Young, R. ( 1984). Negotiation of outcome and negotiation ofmeaning in ESL classroom interaction. TESOL Quarterly, 18, 525-526. Young, R. ( 1986). The acquisition of a verbal repertoire in a second language. Penn Worldng Papers in.Educational Linguistics, 2(1), 85-119. 4I

REFERENCES

Young, R. (1988). Computer-assisted language learning conversations: Negotiating an outcome. CALICO Journal, 5(3), 65-83. Young, R. ( 1991 ). Variation in interlanguage morphology. New York: Peter Lang. Young, R., & Perkins, K. (1991 ). Implications ofnew models ofeducational assessment for theories of second language acquisition. Paper presented at a conference on Theory Construction and Methodology in Second Language Acquisition Research, Michigan State University ,East Lansing, October 4-6. Zuengler, J. (1989a). Assessing an interaction-based paradigm: Ho;w accommodative should we be? In M. R. Eisenstein (Ed.), The dynamic interlanguage: Empirical studies in second language variation (pp. 49-68). New York: Plenum. Zuengler, J. (1989b). Performance variation in NS-NNS interactions: Ethnolinguistic difference, or discourse domain? In S. Gass, C. Madden, D. Preston, & L. Selinker (Eds.), Variation in second language acquisition. Volume I: Discourse and pragmatics (pp. 228-244). Clevedon, Avon: Multilingual Matters. Zuengler, J. (1991). Accommodation theory and the contrastive analysis hypothesis: Both better as weak versions than strong. Paper presented at the Second Language Research Forum, February 28-March 3, University of Southern California, Los Angeles.

II

.. •lll

Discourse Variation in First Certificate in English

Discourse Variation in First Certificate in English

Suggest Documents

First Certificate in English - Grimaldi

First Certificate in English - Inlingua

FCE (First Certificate in English) Speaking ... - UsingEnglish.com

FCE (First Certificate in English) Speaking ... - UsingEnglish.com

FCE (First Certificate in English) Speaking ... - UsingEnglish.com

Cambridge First Certificate in English 6 - Assets

FCE (First Certificate in English)- Phrases - UsingEnglish.com

FCE (First Certificate in English) Listening Tactics

FCE (First Certificate in English) Writing Tactics

First Certificate in English - SIOC AS

FIRST CERTIFICATE IN ENGLISH WRITING ... - sek-eso-english

FCE (First Certificate in English) Use of English ... - UsingEnglish.com

Bruno Grimaldi - First Certificate English

FCE (First Certificate in English) Writing- Application Letter Tasks

Grossman The First Certificate in English - University of Birmingham

FCE (First Certificate in English) Listening Tactics - UsingEnglish.com

FCE (First Certificate in English) Speaking- Needs ... - UsingEnglish.com

FCE (First Certificate in English) Speaking Part ... - UsingEnglish.com

FCE (First Certificate in English) Speaking- Mistakes - UsingEnglish.com

FCE (First Certificate in English) Writing- Reviews - UsingEnglish.com

FCE (First Certificate in English) Speaking Part ... - UsingEnglish.com

First Certificate in English (FCE ESOL Cambridge) SAMPLE ...

FCE (First Certificate in English) Writing Tactics - UsingEnglish.com

First Certificate in English - FCE - Speaking-Test - Examiner Comments