Applied Linguistics Table of Contents Volume 21, Issue 4: December 2000.
Anniversary article. Classroom SLA research an...
54 downloads
487 Views
813KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Applied Linguistics Table of Contents Volume 21, Issue 4: December 2000.
Anniversary article. Classroom SLA research and second language teaching PM Lightbown pp. 431-462 Abstract Full-Text PDF (195 KB) Formulaic sequences in second language teaching: principle and practice A Wray pp. 463-489 Abstract Full-Text PDF (196 KB) Testing L2 vocabulary knowledge at a high level: the case of the Euralex French Tests P Bogaards pp. 490-516 Abstract Full-Text PDF (148 KB) Investigating instrument-based pragmatic variability: effects of enhancing discourse completion tests K Billmyer and M Varghese pp. 517-552 Abstract Full-Text PDF (181 KB) Metaphor awareness and vocabulary retention F Boers pp. 553-571 Abstract Full-Text PDF (117 KB) Review. Reading Science: Critical and Functional Perspectives on Discourses of Science. JR Martin, R Veel (eds) Z Kurtán pp. 572-575 Details Full-Text PDF (54 KB) Review. Social Psychological Perspectives on Second Language Learning. RK Agnihotri, AL Khanna, I Sachdev (eds). Contemporary Approaches to Second Language Acquisition in Context. V Regan (ed) M Lacorte
pp. 576-581 Details Full-Text PDF (54 KB) Review. Adjective Intensification - Learners Versus Native Speakers. A Corpus Study of Argumentative Writing. GR Lorenz C Paradis pp. 581-585 Details Full-Text PDF (52 KB) Review. Other Floors, Other Voices: A Textography of a Small University Building. JM Swales A Davies pp. 585-590 Details Full-Text PDF (55 KB)
Applied Linguistics 21/4: 431±462
# Oxford University Press 2000
Anniversary Article Classroom SLA Research and Second Language Teaching Patsy M. Lightbown Concordia University, Montreal In Lightbown (1985a), I summarized SLA research by stating ten generalizations which were consistent with the research to that date. I concluded that SLA research could not serve as the basis for telling teachers what to teach or how. One of the reasons for that was the limited scope of SLA research at that time. Another reason was that most of the research had not been designed to answer pedagogical questions. However, I suggested that SLA research was one important source of information which would help teachers set appropriate expectations for themselves and their students. In this paper, following a review of language teaching practices of the past ®fty years, I reassess the ten generalizations in light of the considerable amount of classroom-based SLA research which has been carried out since 1985, especially that which has addressed pedagogical concerns in primary and secondary school foreign and second language classes. For the most part, this research tends to add further support to the generalizations, and this gives them greater pedagogical relevance. Nevertheless, I argue that teachers need to continue to draw on many other kinds of knowledge and experience in determining the teaching practices which are appropriate for their classrooms.
In 1985 Applied Linguistics published a paper which I wrote about the relationships between second language acquisition (SLA) research and second language teaching (Lightbown 1985a). In that paper, titled `Great Expectations', I observed that, while many of the changes which had occurred in second language teaching were compatible with SLA research, most of those changes predated and/or were quite independent of the SLA research which provided empirical support for them. SLA research was still a young ®eld, and I argued that it was not yet possible to assume that it had progressed far enough to be used as a major source of guidance for either the `how' or the `what' of second language teaching. I suggested that the proper role for SLA research in teacher education was to help in setting realistic expectations for what language teachers and learners could accomplish in a second/foreign language classroom. SLA research in areas such as orders of acquisition, crosslinguistic transfer eects, and age factors could potentially explain why some things were so dicult, in spite of eort and good will on the part of both teachers and learners. I outlined ten `generalizations' which I believed to be consistent with what SLA research had turned up to that point (see Figure 1).
432 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
Ten Generalizations from SLA Research 1. Adults and adolescents can `acquire' a second language. 2. The learner creates a systematic interlanguage which is often characterized by the same systematic errors as the child learning the same language as the ®rst language, as well as others which appear to be based on the learner's own native language. 3. There are predictable sequences in L2 acquisition such that certain structures have to be acquired before others can be integrated. 4. Practice (does not make perfect. 5. Knowing a language rule does not mean one will be able to use it in communicative interaction. 6. Isolated explicit error correction is usually ineective in changing language behaviour. 7. For most adult learners, acquisition stops . . . before the learner has achieved native-like mastery of the target language. 8. One cannot achieve native-like (or near native-like) command of a second language in one hour a day. 9. The learner's task is enormous because language is enormously complex. 10 A learner's ability to understand language in a meaningful context exceeds his/her ability to comprehend decontextualized language and to produce language of comparable complexity and accuracy.
Figure 1: Ten Generalizations from SLA Research (from Lightbown 1985a) In discussing the possible applications of SLA research ®ndings to the practice of foreign/second language (FL/SL) teaching, I endorsed Evelyn Hatch's (1978) reminder that research ®ndings should be `applied with caution' and that the only research question a researcher should answer is the one that he or she had asked in designing the research. I suggested that `only research which is pedagogically based and which asks pedagogical questions can be expected . . . to answer pedagogical questions' (1985a: 183). Few of the SLA researchers whose work was reviewed in 1985 were asking questions about SLA in the classroom. However, in the years since the publication of that paper, the number of studies designed to ask pedagogical questions has increased dramatically. In this paper, I want to look at some of the classroombased SLA research in order to revisit the question of how SLA research might in¯uence second language teaching.1
PATSY M. LIGHTBOWN
433
CHANGES IN APPROACHES TO FL/SL TEACHING In 1998, I was teaching a course in a university teacher training program in Quebec. Readings for the course covered a variety of issues related to the impact of instruction and feedback on second language acquisition (SLA) in the classroom. Early in the term, the students read Lightbown and Spada (1990), a paper based on an analysis of the classroom practices of four teachers of English as a second language (ESL) in Quebec. One of the students opened the discussion of the paper with the following comment and question: `You're going to hate me for this for the rest of the term but I just have to ask. I really thought this article was very interesting and I agree with what you said. But do you really have to do research to show that it is better for students to get a mixture of communicative interaction and form-focussed instruction? Doesn't everybody know that?' At the end of the 1990s, this student and the other pre-service teachers in my class assumed that, as second language teachers, they would be responsible for creating opportunities for their students to hear and read interesting and motivating samples of the target language and to interact in group- and pairwork. In addition, they anticipated that they would sometimes need to provide form-focussed instruction and feedback to help students make progress on some particularly dicult features in the language they were learning. These were student-teachers whose training had convinced them that second language classrooms should be characterized by a variety of activities, with an emphasis on those which engage students in meaningful interaction, but with an awareness on the part of the teacher that some attention to language form is also necessary. I was struck by the contrast between the assumptions of these new teachers and those of both pre-service and experienced teachers who had been in my classes in earlier years. Those teachers had been persuaded of the validity of dierent pedagogical practicesÐranging over the years from strict step-by-step drill to open-ended communicative interaction. The student who asked me about the usefulness of stating the obvious gave me a reason to take the class through some recent history of language teaching and research on second language teaching in the classroom. I could have gone back `25 centuries' (Kelly 1969), but I stayed with what I had experienced in my own lifetime as a learner, teacher, and researcher in classroom second language acquisition. Like the student in my class, a layperson, not familiar with research and practice in second language teaching and learning, would very likely also ®nd it odd that researchers would feel compelled to seek empirical support for the notion that learners could bene®t from both focussed instruction and the freedom to use the language communicatively. It is only in the context of understanding how `experts' have viewed the process of FL/SL teaching and learning that the conclusion requires explanation. It is important to keep in mind that what `everybody knows' is not always consistent with the facts. In the 1950s and 1960s, when I was a student in high school and university,
434 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
`everybody knew' that foreign languages were learned through reading and translating literary texts, learning grammar rules, and memorizing vocabulary lists. In North America, that type of teaching and learning was challenged by the advent of audio-lingual teaching, and by the time I began teaching highschool French in the late 1960s, `everybody knew' that language was learned through imitation, repetition, and drill, leading to the formation of habits of correct language production (Brooks 1964; Lado 1964). In the 1970s, when I began to work as a teacher trainer and classroom SLA researcher in Quebec, the challenge I faced when talking with experienced or recently trained teachers was to convince them that it was all right to relinquish some control of their structure-by-structure audio-lingual lessons. Even though there were many articles in the North American theoretical literature which questioned the role of audio-lingual FL/SL teaching (see, e.g., Carroll 1966; Chastain 1971), teachers who, like me, had been trained in the late 1960s saw their responsibility as making sure that students practised and overlearned every dialogue and drill. We had been taught that we should not allow students to venture into `free' use of the language for fear that unstructured language production would inevitably lead to errors, interfering with the establishment of correct habits. At that time, there was little available SLA research, and the `Language Acquisition' courses I taught made reference mainly to research in child language acquisition. I taught my students about the growing evidence that the model of language acquisition which underlay audio-lingual teaching had been challenged as inadequate to explain child language acquisition (e.g., Miller 1965). The SLA research which I could ®nd seemed of little relevance to my students. Much of the American SLA research published in the 1970s dealt with university students or other learnersÐboth adults and childrenÐwhose exposure to the second language included a signi®cant amount of informal as well as instructed learning. Most of the preservice and in-service teachers in my class were oriented to the teaching of English in the primary and secondary schools of Quebec, a setting which is more like English as a foreign language in Europe than like ESL teaching in the United States. I made frequent appeals to Savignon's (1972) study of university students learning French as a foreign language. There were few studies of classroom foreign language learners of elementary or school age and few studies of learners whose principal exposure to the language was in the classroom. In the absence of such studies, students found their Language Acquisition courses `interesting' but not very `practical'. The hypothetical similarity of FL/SL acquisition to child language development (see, e.g., Corder 1967) remained to be con®rmed. Meanwhile, teachers were more at ease with their textbooks based on Lado's (1964) `scienti®c approach' to FL/SL teaching. It was clear that there was a need for classroom-based SLA research to determine the extent to which L1 research and non-classroom SLA research ®ndings had relevance to classroom teaching. By the late 1980s, the pendulum had swung, and communicative language teaching (CLT) had swept through many FL/SL classes, including those in
PATSY M. LIGHTBOWN
435
Quebec. By this time, `everybody knew' that exposure to comprehensible input and the opportunity to interact in group work would give students what was necessary for language acquisition to proceed. The teachers my research group and I observed in the late 1980s often told us that it was not good to point out students' errors or to focus on any single linguistic feature. Rather, it was enough to engage students in interesting and meaningful interaction. The language would `take care of itself '.2 The syllabus developed by the Ministry of Education was based on ideas drawn from the British notional/functional syllabus (see below), but in contrast to the intentions of at least some of the British proposals, the pedagogical approach in Quebec had evolved as a version of CLT which seemed to exclude any focus on language form in the second language classroom. Recently trained teachers and teachers who attended teachers' conferences and workshops were very aware of ideas about language teaching which are associated with Krashen's (1982) views of language acquisition and learning. Within a few more years, these ideas had become so integrated in the language teaching mainstream that many teachers were not aware of their origin. Nevertheless, they were con®dent that any attention to language form would cause students to feel stressed or anxious without really helping them become more pro®cient in the second language. It was in this context of rejection of form-focussed instruction that we had `stated the obvious' in Lightbown and Spada (1990). Even though the in¯uence of SLA research was clearly reinforcing the shift to CLT in Quebec in the 1980s, the original in¯uence for this change was not SLA research but rather the notional-functional syllabus movement in Great Britain (see, e.g., Wilkins 1976). In Great Britain, and throughout the world where British teachers taught English as a foreign language, the emergence of CLT was very much in¯uenced by the work of British applied linguists such as Breen, Brum®t, Candlin, Widdowson, and Wilkins. The reference lists of articles and books on language teaching published in Britain in the 1970s include such names as Halliday, Hymes, Lyons, and Searle. The case for approaches to language teaching that were eventually subsumed under the name of CLT was not based on SLA. It came partly from a debate within linguistic theory and partly from discussions among language teachers and textbook writers about what students really needed and wanted to do with the language they were trying to learn and whether strictly structure-based approaches could lead to the accomplishment of their goals. The linguists and applied linguists who had such a profound in¯uence on the development of several varieties of CLT were dissatis®ed with pedagogical approaches which treated language as a set of morpho-syntactic rules rather than as a richer system of communication involving many kinds of competence other than the purely syntactic. For the developers of new approaches, CLT re¯ected a move away from linguistics as the main or only basis for deciding what the units of language teaching would be. They wanted more emphasis on language in context, language in use for a purpose, language to do something with, not just language form for its own sake. Their concerns had to do with authenticityÐ
436 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
teaching real language rather than contrived examples of grammatical rules. They also anticipated that more communicative interaction would have a positive eect on learners' motivation. In Europe, advocates of notionalfunctional syllabuses argued that a syllabus might be ordered in terms of the uses to which language would be put (e.g., Munby 1978). Thus, it was proposed that rather than starting from `the verb to be in simple sentences', a syllabus might begin with greetings and introductions and progress through invitations, expressions of thanks, apologies, compliments, reporting and seeking information, etc. Students who were learning languages for speci®c purposes would need vocabulary or structures which were found to be appropriate for their ®elds of study or work. This approach was debated among the proponents of CLT who argued that this was simply another way of breaking language down into isolated units and urged that language be taught in ways which better re¯ected the reality of complex discourse in context (Widdowson 1978). Nevertheless, the primary cause of change in language teaching in Great Britain was not the ®ndings of language acquisition research, but rather the re¯ections of educators and applied linguists who wanted to see a greater emphasis on the rich complexity of language use (Mitchell 2000). To be sure, one of the reasons that SLA research was not referred to in the British language teaching literature was that there was so little of it! Some researchers expressed an interest in the emerging research about child language acquisition, and there was an expectation that SLA research might eventually play a role (e.g., Corder 1967; Wilkins 1974). Early attempts to propose pedagogical practices on the basis of the very ®rst SLA research ®ndings brought forth several cautionary warnings (e.g., Hatch 1978; Tarone et al. 1976). SLA research exploded in the late 1970s, and in 1984, Littlewood produced a valuable book which gave teachers and teacher trainers an optimistic but cautious overview of SLA research and its possible relationship to teaching. Approaches to language teaching changed in the United States as well. The changes were in¯uenced by cognitive psychology and by the sea change in linguistic theory which followed Chomsky's challenge to Skinner's view of verbal learning. It also grew out of teachers' frustrations with the implementation of audio-lingual teaching in contexts for which they felt it was not well-suited. Critics of the audio-lingual approach to teaching argued that second language learners, especially older children and adults, could use their general cognitive and problem-solving abilities in learning a second language, and that the rote memorization and drill approach failed to capitalize on these abilities (Ausubel 1964; Chastain 1971). As SLA research studies began to appear with rapidly increasing frequency in the 1970s, the compatibility of SLA ®ndings with the independently motivated rejection of audio-lingual teaching tended to strengthen the move toward more `communicative' or `cognitive' approaches to language teaching (Brown 1980).
PATSY M. LIGHTBOWN
437
Another signi®cant development in FL/SL teaching which predates SLA research is the Canadian experience with French immersion. This experiment in second language teaching was motivated by parents' dissatisfaction with traditional language teaching, and it began in the mid-1960s, well before there was a body of SLA research to support the idea of comprehension-based or content-based language teaching. In the very ®rst widely circulated reports on French immersion, students were said to have excellent listening comprehension and the ability to use French to learn subject matter. It was reported that their spoken French still contained many errors, even after several years of immersion, but the quality of students' spoken French was not at the top of the list of topics given close scrutiny for the ®rst few years of French immersion programs. Concern about oral production ability was overshadowed by issues such as subject matter learning, mother tongue maintenance, and personal identity, and on these measures, the results were overwhelmingly reassuring (Lambert and Tucker 1972). In the late 1970s and increasingly in the 1980s, SLA researchers began to look at French immersion classes as a valuable context for studying aspects of second language acquisition. The success of French immersion programs, together with evidence from SLA research and theory of the early 1980s, led to the development of content-based and `sheltered' FL/SL courses in the 1980s (e.g., Ready and Wesche 1992). However, by the mid-1980s, researchers were raising questions about the adequacy of a purely comprehension- or contentbased approach to language teaching for achieving high levels of pro®ciency in language production (Harley and Swain 1984; Swain 1985, 1988). In summary, FL/SL teaching changed dramatically in the years between 1960 and 1985, but the changes were due largely to factors other than ®ndings from research which was carried out in classrooms or which was designed to answer pedagogical questions. As I noted in 1985, even `the natural approach' (Krashen and Terrell 1983)Ðthe approach which is most closely related to SLA research and theory of the 1970sÐwas ®rst developed out of Terrell's experience as a teacher (Terrell 1977) and later reinforced by Krashen's interpretation of the early research in SLA (Terrell 1982). Language teachers and curriculum designers felt the need for change, but although the changes were sometimes in¯uenced or reinforced by the SLA research which had been reported at the time, they could not have been based mainly on classroom SLA research because so little SLA research had been done in classroom contexts up to that point.
SLA RESEARCH IN THE CLASSROOM In the year 2000, the scope and volume of SLA research are quite dierent from that which I summarized in 1985. At the end of that paper I noted that SLA researchers whose work was focussed on solving theoretical puzzles were increasingly separating their research activities from those of researchers whose questions were more pedagogical in nature. That trend has continued,
438 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
and there are now several conferences and journals where SLA research is viewed entirely within one theoretical framework or another. At the same time, there has been a huge increase in SLA research which is either carried out in the classroom or which has been designed to answer questions related to FL/SL pedagogy. The research orientations range from classroom ethnographies to experimental interventions to test hypotheses about feedback techniques. Since 1985, numerous books (see the Appendix for a partial list) and hundreds of articles, theses, and conference papers have focussed on research in FL/SL classrooms and on the relationship between SLA research and FL/SL learning in the classroom. The speci®c goals of the various research projects dier, but there is a unifying desire to identify and better understand the roles of the dierent participants in classroom interaction, the impact that certain types of instruction may have on FL/SL learning, and the factors which promote or inhibit learning. We are a long way from ®nding answers to all these questions, and it is clear that no single type of research can address them all. It is important not only that we continue to do classroom-based research but also, as Van Lier urges, that the research `be open rather than closed, that is, [that it be] formulated and designed in such a way that it can be used to maintain a dialogue between [interpretive and experimental] types of research' (1988: xiv). The dialogue needs to include not only researchers with dierent orientations but also teachers and students whose experiences we seek to understand and, eventually, to enhance. In 2000, it is no longer correct to say that changes in language teaching have not been in¯uenced by SLA research. The in¯uence of SLA research is now evident in textbooks and teacher training programs and in proposals for curriculum design (see, e.g., Long and Crookes 1992). Some aspects of SLA research and theory have become so completely integrated into mainstream FL/SL pedagogy that they are referred to without reference to their sources. Notions such as the silent period, focus on form, or developmental stages are often taken for granted by teachers trained since the 1980s. There remain many unanswered questions about the relationship between SLA research and the classroom. Researchers and practitioners still have good reason to ask whether, and if so how, SLA research should in¯uence pedagogy. As Crookes says, `If the relationship were simple, or not a source of concern, I do not think it would come up so often' (1998: 6).
SLA RESEARCH GENERALIZATIONS (1985) REVISITED I would like now to review the generalizations I proposed in 1985, looking at the extent to which SLA research which has been carried out in the classroom provides support for the relevance of these generalizations to classroom learning or raises questions for further research. Each generalization is accompanied by references to some related classroom-based studies related to the generalization, with special emphasis on research carried out in primary and secondary FL/SL classrooms.
PATSY M. LIGHTBOWN
439
1. Adults and adolescents can `acquire' a second language In the original 1985 discussion of this generalization, the term `acquire' was used to acknowledge the distinction that Krashen (1982 and elsewhere) made between linguistic abilities which learners develop in the absence of metalinguistic instruction (acquisition) and what they come to know about the language through formal instruction or metalinguistic analysis (learning). Classroom research has provided additional support for the conclusion that some linguistic features are acquired `incidentally'Ðwithout intentional eort or pedagogical guidance. While there is wide agreement that learners in both classroom and informal learning environments come to know things that were never the subject of explicit teaching, there are dierent views about the extent to which this really takes place without the learner's `awareness'. Schmidt (1990, 1994) has been particularly in¯uential in pointing out the dierence between performance which appears to be automatic and the acquisition of the knowledge which underlies that performance. He has argued that, in order for development to take place, learners must `notice' the dierence between their interlanguage and the target language. The nature of such noticing is the topic of considerable ongoing debate, but all would agree that it cannot be equated with (or limited to) the content of the formal instruction to which learners have been exposed. Classroom research in French immersion contexts provides some evidence for learners' ability to acquire language in the absence of direct instruction of particular linguistic features. In these classes, young English-speaking students, whose attention was focussed on following classroom instructions and learning the subject matter of their school program, also acquired the ability to understand both written and spoken French and to produce it with a considerable degree of ¯uency and con®dence (Swain 1991). Similarly, in intensive ESL courses with 11- and 12-year-old francophone children in Quebec, where teachers engaged students in a variety of communicative interactions with little or no focus on language itself, students acquired the ability to use English with ease in simple communicative situations (Lightbown and Spada 1994). There have been many studies of group work and the interaction between learners in pairs in classrooms or in simulated classroom contexts. This research has shown that learners (1) are able to give each other FL/SL input and opportunities for interaction, (2) do not necessarily produce more errors than when they are interacting with the teacher, (3) can provide each other with feedback on error, in the form of clari®cation requests and negotiation for meaning, and (4) bene®t from the opportunity for more one-to-one conversation than they can get in a teacher-centered whole class environment (Gass and Varonis 1994; Long and Porter 1985; Pica 1987; Yule and Macdonald 1990). Most of this research has involved adult learners, but Oliver (1995) has shown that even young children are successful in using interaction to get appropriate input. Kowal and Swain (1994) have found that
440 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
adolescents are able to bene®t from pair work activities in which students work together to reconstruct dictated texts. Much of the research on group and pair work has been carried out in the framework of Long's (1985) hypotheses about the importance of modi®ed interaction in second language acquisition. Swain (1999) and others (e.g., Donato 1994) have used Vygotsky's theory of the role of social learning to explain how interaction contributes to language development. Krashen's hypothesis that comprehensible input is the one necessary building block for language acquisition has also been explored in a number of studies. Elley (1991) reviewed a number of studies, including his own, in which young second language learners bene®tted from reading or being read to in the target language. The ®ndings consistently showed that students who had access to more reading activities learned more of their second language than students in audio-lingual instruction. Krashen and his students have carried out studies of the impact of extensive reading as a source of comprehensible input (e.g., Krashen 1989; Pilgreen and Krashen 1993). These studies are said to show strong support for the comprehensible input hypothesis, especially that corollary of the hypothesis which suggests that the best form of comprehensible input is `reading for pleasure'. However, there is some evidence that learners bene®t even more when their reading is supplemented by interaction with a teacher (Elley 1989). Zimmerman (1997) and others have found that adult students who received some guided instruction in addition to extensive reading acquired more vocabulary than those who had only extensive reading. Lightbown (1992) reported on an ESL program in New Brunswick, Canada, in which young (grades 3 to 5) learners of English as a second language had their entire instructional experience in the form of listening/reading sessions. The course was designed so that students spent their half-hour classes with a book which they had chosen from a large collection displayed in the classroom. Each book had an accompanying audio tape that students listened to using earphones attached to individual tape players on their desks. Teachers were available mainly to help students with technical problems, not to `teach' in any formal sense of the term.3 At the end of three years of this program, the students in the comprehension-based group performed as well as or better than students in a more traditional, modi®ed audio-lingual program. Three years later, however, students who had continued to participate in this type of comprehension-based program did not perform as well on most measures as students who had had the guidance of a teacher and more opportunities for oral and written production tasks (Lightbown et al., in preparation).
PATSY M. LIGHTBOWN
441
2. The learner creates a systematic interlanguage which is often characterized by the same systematic errors as the child learning the same language as a ®rst language, as well as others which appear to be based on the learner's own native language This generalization is at the heart of modern SLA research. Corder (1967) suggested that learners' errors provided insight into the system underlying second language learners' language use, and in the years since that time, innumerable studies have con®rmed that learners develop an `interlanguage' (Selinker 1972) which has systematic properties that are not explained in any simple way by the input learners have been exposed to. Even when students are exposed to the L2 primarily in structure-based classes, they create their own systematic interlanguage patterns which do not simply re¯ect the patterns they have been taught (Lightbown 1991). A number of researchers emphasized the fact that a learner's ®rst language was not the only in¯uence on the shape of the interlanguage and emphasized the importance of looking at actual samples of learner language rather than at contrastive analyses which predicted what learners would ®nd easy or dicult (e.g., Richards 1973.) Some researchers took quite radical positions in rejecting a role for the L1.4 It soon became evident, however, that the role of the ®rst language needed to be reintegrated into interlanguage studies, and L1 in¯uence regained its place in the mainstream of SLA research (see e.g., Gass and Selinker 1983; Kellerman and Sharwood Smith 1986; Zobl 1980a, b). Research with young classroom learners has shown how subtly the ®rst language can aect both learners' production and their perceptions of what is grammatical in the target language. In a study of learners' production of written French, Harley and King (1989) found that the verbs used by Englishspeaking children in French immersion re¯ected the English pattern for showing direction of motion. In English, prepositions or adverbials are used to indicate the direction of the motion expressed by the verb, while, in French, the direction of motion is often included in the verb itself. English learners of French tended to use expressions such as aller en bas (go down) in contexts where French speakers themselves preferred descendre. These eects of L1 are not limited to production of the language. Spada and Lightbown (1999) found that, on a sentence correction task, French-speaking students learning English readily accepted questions such as `Can you play outside?' and `Do you like pepperoni pizza?' while, at the same time, they rejected questions such as `Can the children speak Spanish?' and `What is your father doing?'. Although their interlanguage included `inversion in questions', they appeared to have brought over from French the prohibition on inversion with full nouns (Peux-tu venir chez moi? *Peut-Pierre venir chez moi?) (see Zobl 1979). Even when questions with pronoun subjects were quite well-established in both oral and written production, they continued to reject
442 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
the use of inversion with full noun phrases or proper names. This accounted for their apparently dierent levels of performance on dierent tasks. That is, to the extent that the task allowed or required questions with pronoun subjects, they were able to perform with a high level of target-like accuracy. When noun subjects occurred, however, they gave the impression of being `less advanced'. In fact, the behaviour was quite systematic and re¯ected their own understanding of the constraints on English quite well.
3. There are predictable sequences in L2 acquisition such that certain structures have to be acquired before others can be integrated This generalization is related to the previous one, and classroom research since 1985 has added support to the early SLA research evidence that many linguistic features are acquired according to a `developmental sequence' and that, although learners' progress through a sequence may be speeded up by form-focussed instruction, the sequence which they follow is not substantially altered by instruction (e.g., Ellis 1989). When classroom input is very restricted or when learners' production consists largely of memorized formulas, there is sometimes the appearance of dierence (Weinert 1987). Sometimes, the restricted or distorted samples of the target language which learners are exposed to can contribute a developmental path which re¯ects the acquisition of something other than the target language (Lightbown 1985b). Nevertheless, when learners have adequate opportunities to understand and use their second language, they show considerable similarity in the acquisition sequences. One important implication of this generalization is that we cannot rely on the assessment of students' target-like use of the second language as the only evidence for growth or change in their interlanguage knowledge (see BleyVroman (1983) for an early warning about this practice). The research on developmental sequences makes it clear that progress in a learner's interlanguage will not necessarily show up as greater accuracy. That is, the developmental stages through which learners pass on their way to higher levels of pro®ciency include stages in which their performance is still far from target-like. For example, early stage learners may ask a few formulaic questions correctly (What's your name? What do you want?) before they progress to asking incorrect questions which are original and re¯ect their developmental stage (`What your brother's name?' `Why the children want to go there?') The latter questions are less `accurate' than the former because they do not conform to the target language. But they re¯ect the learner's ability to create new questions, placing the wh-word at the front of a declarative sentence to form a question. This change shows developmental progress beyond the memorized formulas of the earlier stage. Pienemann's (1985) `teachability hypothesis' grew out of earlier research (Meisel et al. 1981) in which developmental sequences were identi®ed in the
PATSY M. LIGHTBOWN
443
acquisition of German word order by learners who were getting no classroom instruction. This hypothesis suggested ways of operationalizing an idea that teachers have always intuitively held: that learners can be taught only what they are ready to learn. Pienemann (1988, 1999) himself tested this hypothesis with learners of dierent ages and in dierent learning environmentsÐboth informal and instructed. He observed that instruction was most eective when it re¯ected the stage just beyond the learners' current stage of interlanguage. In apparent contrast, Spada and Lightbown (1993, 1999) found that communicative input which included many examples of forms typical of a particular developmental stage in the formation of English questions did not appear to be more eective for learners at the `right' stage. Nevertheless, when learners made progress, they moved forward according to the sequence of English question development proposed by Pienemann et al. (1988).5 Keenan and Comrie's (1977) accessibility hierarchy for relative clauses was developed in a typological study of the languages of the world, but it has inspired considerable classroom-based research. Gass (1982) was the ®rst to experiment with the idea that teaching the most marked structures in a hierarchical sequence of the relative clauses would lead learners to acquire the least marked structures in that hierarchy without explicit instruction. Numerous published studies followed Gass's lead showing that instruction could speed learners along in their acquisition of certain structures, but that the sequence of acquisition is not changed by the instruction (see Hamilton 1994 for a review). Most of this research has been done with university-level learners, but Ammar (1996) obtained similar results with young secondary school learners of English as a foreign language in Tunisia. Teachers are often fascinated by research on developmental sequences and wonder whether they should plan their lessons in terms of these patterns. In my view, there are many reasons why such a proposal is neither feasible nor desirable (see Lightbown 1998 for discussion). However, the developmental sequences research should help teachers identify progress in ways other than an increase in target form accuracy.
4. Practice does not make perfect In stating this generalization in 1985, I was thinking of practice as it was often de®ned in audio-lingual classrooms, practice which was characterized as rotelearning outside meaningful language use. When `practice' is de®ned as opportunities for meaningful language use (both receptive and productive) and for thoughtful, eortful practice of dicult linguistic features, then the role of practice is clearly bene®cial and even essential. Nevertheless, research evidence shows that communicative practice in the classroom, as valuable as it is, is not sucient to lead learners to a high degree of ¯uency and accuracy in all aspects of second language production. The original generalization was related to the emerging SLA research on
444 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
developmental sequences. As noted under generalization 3, when learners drill and memorize language material which is beyond their current level of development, they may eventually exhibit `U-shaped' behaviour (Kellerman 1985). That is, their apparently high level of accuracy, based on the use of memorized chunks, suddenly drops and then rises again as they come to create novel sentences. Even when instruction is not oriented to the rote learning of whole phrases or sentences, learners may have diculty recognizing the components which make up the chunks of language they are frequently exposed to. For example, Harley (1993) has shown that French immersion students cannot always distinguish between je, the ®rst person singular pronoun, and j'ai, the pronoun plus the verb `have'. When students produce sentences such as J'ai aime cËa (which would be translated word-forword as `I-have like that'), they are not moving in a straight line toward perfection because they will have to unlearn, or at least reanalyze, these sentences. Nevertheless, several researchers have pointed out the positive role for formulaic material in second language acquisition, both in and out of classroom contexts (Mitchell and Martin 1997). Myles et al. (1998, 1999) have carried out an extensive study of the use of language chunks which learners have practised in French as a foreign language classes in Britain. They examined the way in which learners use rote-learned material outside the situations in which it was originally taught. They showed how learners used the complex material for its communicative value while at the same time they were beginning to use the less target-like language which was more typical of their developmental level. The extent to which learners use the complex, memorized chunks for language acquisition by breaking them down for analysis remains controversial, but it seems very likely that the communicative eectiveness and the con®dence-building associated with the ability to produce longer, albeit formulaic, utterances can play a role in maintaining learners' motivation. Furthermore, current research which adopts a connectionist or emergentist perspective on language acquisition suggests that formulaic chunks or high frequency collocations account for a great deal more of our ¯uent language use than is re¯ected in some recent linguistic theories (Ellis 1996; Wray 1999).
5. Knowing a language rule does not mean one will be able to use it in communicative interaction This generalization is one which has been apparent to language teachers and language learners for as long as languages have been taught. Teachers have always observed that students can do well on a test but still fail to use the same features when they engage in spontaneous conversation. Students complain that, after years of language study, they still can't `order breakfast'. Krashen (1982) went further and argued that the kind of knowledge which learners get when they learn metalinguistic rules will be of little use to them in natural communication unless they have time to `monitor' their
PATSY M. LIGHTBOWN
445
performance. That is, in his view, the metalinguistic information does not have any direct eect on what he called `acquired' knowledge. Ellis (1993) has adopted what he calls a `weak interface' position, suggesting that instruction draws learners' attention to language features and permits them to develop knowledge of those features if they are developmentally ready to do so (see also Lightbown 1998). A considerable amount of classroom research in SLA has focussed on the role of instruction in helping learners to move beyond their current interlanguage state. In classroom instruction, Long (1991) makes a distinction between what he calls focus on formsÐdiscrete point, step by step grammar instruction, and focus on formÐcorrective feedback which is fully integrated within ongoing communicative activities. There has been little experimental research which directly compares form-focussed instruction which is provided as the need arises within a communicative interaction (Long's focus on form), and form-focussed instruction in isolated grammar lessons (Long's focus on forms). However, a number of studies have compared learners' language development in CLT without focus on form to that which is achieved in CLT with focus (e.g., Doughty and Varela 1998; Harley 1989; Lightbown and Spada 1990; Lyster 1994). The results have provided strong support for the inclusion of focus on form in the CLT classroom. In fact, there has been enough research on this topic that Norris and Ortega (2000) have been able to conduct a meta-analysis of such studies. They conclude that the research con®rms that instruction which includes focus on form does make a positive dierence for classroom SLA. J. White (1998) tried to improve the ability of young francophone learners of English to learn the English agreement rule for possessive determiners by exposing them to `enhanced input' (Sharwood Smith 1993) in the form of many, many examples of his and her in stories, games, puzzles, and poems. She enhanced the input by using bold letters, underlining, or italics to highlight these forms in the texts students read. She found that students exposed to this type of input did make more developmental progress in their use of possessive determiners than students who were not exposed to such a ¯ood of enhanced input. In more recent research, she has provided learners with more explicit information about the relationship between the determiners and the nouns to which they are attached (J. White 1999). She found that learners who were taught a rule of thumb used possessive determiners at a more advanced stage on an oral communication task than a comparison group of students who had not been taught the rule. She suggests that knowing the rule had enabled the experimental group to notice and incorporate the English way of assigning gender to the possessive determiners. L. White (1991) has argued that second language learners are far more likely than child L1 acquirers to develop grammars which are too general. This appears to be because they draw on features of the L1 as well as input from the L2. This places learners in the position of needing to notice that pro®cient speakers of the L2 are not saying something which the learners are saying.
446 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
This is especially problematic in classes where students share the same L1, of course. White argues that they require `negative evidence' in the form of instruction or corrective feedback. The classroom research comparing input ¯oods with more explicit instructional intervention takes up this issue. For example, Trahey and White (1993) exposed learners to a meaning-focussed ¯ood with very large numbers of sentences with correct adverb placement. This input ¯ood was not eective in getting learners to stop using an interlanguage form which was never present in the input except from other learners. However, corrective feedback and explicit instruction did have this result (White et al. 1991). Other researchers have also found evidence for a relationship between explicit knowledge and language performance that is closer than that which exists in L1 speakers or that which is suggested by Krashen's acquisition/ learning hypothesis (see, e.g., Green and Hecht 1992; Han and Ellis 1998). Nevertheless, there continues to be debate about the extent to which explicit rule knowledge shapes learners' underlying L2 linguistic competence or in¯uences L2 performance in genuinely communicative situations (see Bialystok 1994 and Schwartz 1993 for two theoretical perspectives).
6. Isolated explicit error correction is usually ineective in changing language behaviour I believe that the evidence still supports this generalization. Learners' interlanguage behaviour does not change suddenly when they are told that they have made an error. This does not mean, however, that feedback on error is not bene®cial. The evidence seems to suggest that error feedback can be eective, but it must be sustained over a period of time, and it must be focussed on something which learners are actually capable of learning. In typical CLT or content-based instruction, it may be necessary to provide some explicit indication to learners that feedback is directed to language form instead of or in addition to the meaning of a student's utterance. Lightbown (1991) observed a teacher who oered frequent, often humorous, corrections each time the francophone students in her class used `you have' rather than `there is' as an introducer form in sentences such as `You have (=There's) a boy beside the table.' This error had been observed in the oral production of hundreds of students whom we had observed in similar intensive ESL classes. Unlike students in classes where teachers ignored the error, the students in this teacher's class eventually stopped making the error and, more importantly, they were still using the correct form months later when they were no longer receiving the corrective feedback. One type of feedback that has been the subject of a good deal of SLA research in CLT classes is the `recast'. A recast is an utterance by a teacher or other, usually more pro®cient, speaker which rephrases the utterance of a learner, preserving the original meaning, but correcting the error(s) that occurred in the original utterance (Long and Robinson 1998). Laboratory-based studies and
PATSY M. LIGHTBOWN
447
studies where learners interact one-to-one with a more pro®cient interlocutor have shown evidence for the bene®cial eects of focussed recasts, that is, recasts in which a single linguistic feature is targeted for recasting (Leeman 2000; Long et al. 1998; Mackey and Philp 1998; Oliver 1995). Findings from classroom studies are not as clear. In a descriptive study of corrective feedback in French immersion classes, Lyster has shown how dicult it can be for students in content-based classrooms to distinguish between feedback which con®rms the content of what they have said from feedback which is meant to provide information about linguistic accuracy or pragmatic appropriateness. In immersion and other content-based instruction, teachers feel a primary responsibility to ensure that students learn the subject matter of the courseÐscience, social studies, or mathematics. For this reason, they do not always draw attention to errors in form, as long as the students show that they understand the content. In this meaning-focussed context, Lyster and Ranta (1997) report that recasts were the most frequent form of teacher response to students' sentences containing errors. However, recasts were the least likely type of feedback to lead to uptake (an immediate response to teacher feedback). In further analysis, Lyster (1998a) found that teachers tended to use the same praise markers (e.g., `bravo') and the same proportion of recasting and noncorrective repetition when the content of student utterances was correct content, whether the form was correct or not. It is likely that, without other cues, learners could not tell whether their teacher's feedback was intended as a non-corrective repetition or rephrasing of a correct response, or an indication that, while content was correct, there was an error in the form of the original utterance. A student's assumption that the teacher's feedback was a con®rmation of a correct response was further supported by the fact that teachers did not tend to wait for students' uptake following a recast, and the classroom conversation typically just moved on after the recast. Students were more likely to produce uptake when the teacher made it clear that a change was expected, for example, when the teacher recast only part of the student's utterance and waited for him/her to complete it with the desired form (see also Chaudron 1977). In a further analysis, Lyster (1998b) observed that dierent types of errors (lexical, phonological, grammatical) were aected dierently by dierent types of feedback, and that phonological errors were most likely to receive student uptake following recasts. Like other researchers studying the impact of classroom feedback, Lyster (1998a) has emphasized that `it is unwarranted to equate learner uptake with L2 learning' (p.75). Nevertheless, an immediate response does provide some evidence that learners are noticing the feedback and interpreting it as a focus on the form instead of (or in addition to) the meaning. In an experimental classroom study, Doughty and Varela (1998) used a two-step technique which they called `corrective recasts' in a science class for ESL students at the elementary school level. The teacher in the experimental condition provided students with feedback on their use of past tense forms in
448 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
the context of reports on science experiments. In the `corrective recast' technique, the teacher ®rst repeated a student's erroneous utterance, usually adding emphasis to the incorrect form of the verbs which the student had used. Then, if the student did not spontaneously repair the sentence, the teacher provided the correct form as a recast, and sometimes had students repeat the correct form after it had been provided. This two-step technique gave students clear information about what the teacher wanted them to pay attention to. Students who received this experimental feedback treatment improved in their use of past tense forms, using fewer incorrect forms and showing an increase in the use of both target-like forms and interlanguage forms that they had not used prior to the experiment. The control group students who continued in the regular science classes without such feedback did not show the same improvement. Indeed, on some measures, their performance grew worse. These ®ndings with respect to feedback on error are congruent with Spada's (1997) and Ellis' (1995) reviews of research on the role of form-focussed instruction in SLA. They concluded that the research studies which showed an eect for instruction in the context of CLT were those in which there was an element of explicitness in the instruction. That is, the eects of instruction were observed when students' attention was more directly focussed on the feature to be learned than when the object of learning was not made clear.
7. For most adult learners, acquisition stops . . . before the learner has achieved native-like mastery of the target language 8. One cannot achieve native-like (or near native-like) command of a second language in one hour a day The Critical Period HypothesisÐthat post-puberty learners of a second language will always be distinguishable from learners who have had sustained substantial exposure beginning in early childhoodÐcontinues to ®nd support in research which focusses on the long-term outcomes of earlier and later exposure to second languages (see Long 1990 for a review). The research is not without controversy, however, and some researchers point to ways in which the hypothesis needs to be re®ned (see, for example, Birdsong 1998; Bialystok 1997; Singleton 1989; White and Genesee 1996). The Critical Period Hypothesis is often confused with the notion that `younger is better' in second language acquisition. In many places, children who speak minority languages or languages without political status are placed in second language instructional settings before they have an opportunity to develop literacy skills in their L1. Such practices, based on political imperatives, as well as the belief that young children will easily acquire a second language, sometimes result in poor mastery of either language and in
PATSY M. LIGHTBOWN
449
educational diculties with long-term negative consequences (Cummins 1991; Wong Fillmore 1991). In the context of the foreign language classroom, the relevance of the Critical Period Hypothesis is questionable. The reality is that perfect mastery of a target language is rarely attained, even when learners begin at an early age. There are many reasons for this. One is that learners in a foreign language environment usually have only the teacher as a model of a pro®cient speaker, and even the teacher may not provide a native-like language model. All other input comes from learners like themselves, and learners who hear and understand each other's interlanguage varieties inevitably reinforce some of the non-target aspects of that interlanguage (Lightbown 1985b; see also Wong Fillmore 1991). For many years, classroom-based research has suggested that, in instructional settings, the age at which instruction begins is less important than the intensity of the instruction and the continuation of exposure over a sucient period of time (Burstall 1975; Stern 1983). The most important reason for incomplete acquisition in foreign language classroom settings is probably the lack of time available for contact with the language. Children learning their ®rst language and young children living in a second language environment are in daily contact with the target language community. Through friends as well as schooling, they have thousands of hours of contact with the language. When FL/SL learning is limited to classroom instruction, the number of hours is much more limited. Furthermore, in programs where learners begin learning at an early age and then do not continue with the language, the pro®ciency which was developed may be lost. In instructional settings, if the total amount of time is to be limited, it is likely to be more eective to begin instruction when learners have reached an age at which they can make use of a variety of learning strategies, including their L1 literacy skills, to make the most of that time (Harley and Hart 1997; MunÄoz 1999; Singleton 1989). In addition, a later start may mean that learners will have more opportunity to sustain contact with the language into adolescence and adulthood and to have opportunities to use the language outside the classroom. The advantages of older learners are less apparent in contexts where learners have a great deal of sustained informal exposure outside the classroom (Slavo and Johnson 1995). The intensity of the exposure and the opportunity to continue using the language over a long period of time is as important as the starting age in the eectiveness of classroom instruction. Our research in Quebec has shown that students who have intensive exposure to the second language near the end of elementary school have an advantage over those whose instruction was thinly spread out over a longer period of time. That is, even though students began at the same age and received a comparable number of hours of instruction, the more compact instruction was more eective (Collins et al. 1999; Spada and Lightbown 1989). Students whose exposure to the language was sustained into high school, through enriched ESL courses or through contact
450 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
with the language outside of school maintained this advantage (Dussault 1997; Lightbown and Spada 1991). Similar results have been observed in French immersion (Genesee 1987; Turnbull et al. 1998).
9. The learner's task is enormous because language is enormously complex In stating this generalization in 1985, I was thinking mainly in terms of the morphosyntactic complexity of language. The challenge learners face in this regard remains daunting. As noted above, learners need a great deal of time, as well as opportunities for exposure to language in a variety of contexts, before they can master its many subtleties. Many students never reach that mastery of the morphosyntax or of the lexicon. In addition, students in foreign language learning environments face a particular challenge because their classroom exposure to the language leaves them without adequate opportunities to learn appropriate pragmatic and sociolinguistic features of the language (Lyster 1994; Bardovi-Harlig and DoÈrnyei 1998). Tarone and Swain (1995) point out that students in French immersion settings who do not have out-of-school exposure to French fail to learn the informal register appropriate for use with adolescent francophones who would be their peers. Lyster (1994) found that they also had few opportunities to learn the highly formal register appropriate for dealing with adult strangers. Speci®cally, they did not have opportunities to hear markers of formality and politeness such as vous as a respectful second person singular pronoun and conditional form of verbs in a French immersion class.6
10. A learner's ability to understand language in a meaningful context exceeds his/her ability to comprehend decontextualized language and to produce language of comparable complexity and accuracy There is plenty of evidence that learners are able to get the meaning from the language they hear, even if they do not understand all of the linguistic features which contribute to making the meaning. They do this by using contextual cues and world knowledge. This is a very positive factor in the success of CLT and content-based language teaching. However, ongoing research has not tended to con®rm Krashen's hypothesis that exposure to comprehensible input will lead learners to high levels of pro®ciency without some pedagogical guidance. As Swain (1985, 1995) and others have pointed out, learners need more and dierent language knowledge in order to produce language which is as complex and accurate as the language which they appear to `understand'. A number of researchers have observed that there are some features of language which either develop quite slowly, or never seem to develop fully, in learners who are exposed to the language in contexts where the emphasis is
PATSY M. LIGHTBOWN
451
exclusively on getting the meaning and never (or almost never) on learning speci®c linguistic features. This may be due to the very low frequency of some linguistic forms in classroom interaction or to other limitations inherent in the types of interactions which occur in classrooms. For example, White (1991) found that sentences containing adverbs were rare in the classroom language to which francophone students were exposed in Quebec ESL classes. Swain (1988) found that teachers in French immersion classes often used the historical present or future (e.g., `How do you think that these plantations are going to change life in the Antilles?') while teaching history lessons, thereby reducing the frequency with which students were exposed to past tense forms in contexts referring to past events (Swain 1988: 71. Translation from French is in the original). Even when forms are frequently present in classroom input, learners may ®lter them out because of characteristics of their L1 or their current interlanguage. Correct use of gender forms is notoriously dicult for students in French immersion and yet virtually every time a noun is used in French, the sentence contains one or more indicators of the grammatical gender (see Harley 1998). Francophone students in intensive ESL classes fail to produce English plurals correctly, even when they are fairly advanced. One ESL student insisted to me that `English people don't always add the -s.' The plural -s exists in written French, but it is not pronounced in the oral language. Apparently this student `heard' unpronounced plurals in oral English. In their research, VanPatten and his colleagues create situations in which understanding the meaning of a sentence requires learners to focus on the speci®c linguistic features which, in other situations, they overlooked because they could interpret the sentences with the help of contextual clues. VanPatten and Cadierno (1993) observed that university students of Spanish as a foreign language used contextual clues and thus gave the appearance of understanding the ¯exible word order rules of Spanish. However, when they had to depend on the language alone, they made errors based on their lack of knowledge of word order. For example, they interpreted Lo sigue el gato as meaning `It [the dog] follows the cat' rather than the correct meaning `The cat follows it [the dog]'. In this misreading of the sentence, students were showing that the clue in the form of the [object] pronoun was less powerful than their expectation, carried from English, their L1, and con®rmed in much Spanish input, that sentences would follow subject-verb-object order. VanPatten and Oikkenon (1996) replicated the study with high school students and found con®rmation that it was not the explanation which was provided to learners that led to improvement, but rather the opportunity to give their full attention to understanding how the language form encoded meaning. In his research, VanPatten has maintained the hypothesis that the processing of input for meaning is at the heart of language acquisition. He has adapted Krashen's input hypothesis, however, by making a more precise claim about the kind of input that is bene®cial to learners. His pedagogical recommendations are also dierent in that he proposes that the input needs to
452 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
be adapted in very speci®c ways rather than assuming that, in communicative situations, learners will ®nd the input they need. In educational institutions where second language learners must learn both their second language and the subject matter of the school, researchers have found that there can be a very long period between the development of an ability to understand language in social contexts and the ability to understand complex ideas which are conveyed in the second language. Collier (1989), Cummins (1984), Hus (1997), and others have shown that after years of education in a second language environment, second language speakers may continue to experience diculty with language which refers to complex cognitive/academic contexts and in situations where the meaning cannot be derived from contextual cues or prior knowledge. In these situations, gaps in their knowledge of the language become apparent. Lightbown and Spada (in press) report on interviews in which teachers in Northern Quebec were asked about the diculties Inuktitut-speaking learners were having with doing their school work in a second language. Nearly all teachers felt that the students' diculties were largely due to their lack of mastery of the second language. In that same study, we asked students to perform tasks which forced them to rely on language alone to understand or express relatively complex ideas. We found that many high school students, whose education had been entirely in French or English for ®ve or six years (since grade 3), were still struggling with the second language in situations where they could not depend on rich contextual information to make the meaning clear. When we spoke with some school administrators, however, we were assured that students were very comfortable in using the language and that their academic diculties could not be due to a lack of language pro®ciency. Further discussion soon revealed that what these administrators had observed was the students' ability to use the language at the local store or at the hockey arena, not in the context of a history or science lesson.
CONCLUSION Some things have changed a lot since I was working on `Great Expectations' in the early 1980s. The biggest change, perhaps, is the sheer volume of SLA research which has focussed on pedagogical questions. There is no doubt that there is now a rich literature of SLA research which can help shape teachers' expectations for themselves and their students, and provide valuable clues to eective pedagogical practice. Nonetheless, there remain a number of concerns regarding the application of research ®ndings to classroom practice, and Evelyn Hatch's (1978) admonition to `apply with caution' is as pertinent now as it was then. Unfortunately, such caution is not always used. For example, Truscott (1996, 1999) uses SLA research ®ndings to support the recommendation that feedback on error has no place in the FL/SL classroom, and Krashen (1989) claims that research con®rms that pleasure reading will eliminate the need for guided instruction in L2 vocabulary acquisition. Such
PATSY M. LIGHTBOWN
453
recommendations for pedagogical practice concern me for two reasons. First, I don't believe that they are consistent with much of the classroom SLA research cited above. A second concern is that they invite teachers to engage in pedagogical behaviour which is not compatible with their understanding of their role as teachers. This is not to say that anything which goes against teachers' intuitions is incorrect. For one thing, as the review in the ®rst section of this paper makes clear, our pedagogical intuitions are partly shaped by the theories of language acquisition on which our own training was based. Current researchers challenge those theories, and future research is quite likely to challenge the views we hold now. Thus, it is completely appropriate for teachers and researchers to question intuitions about FL/SL pedagogy and to explore their validity. But when researchers make strong claims that are at odds with the views teachers have developed through their experience with learners, and when those claims are made on the basis of research which has been done in contexts which do not re¯ect reality as the teachers know it, they are likely to alienate teachers and lead them to dismiss researchers as ivory tower oddities. It is essential for SLA researchers to enter into dialogue with classroom teachersÐnot only so that teachers can know what researchers are saying, but also so that researchers can hear what teachers are saying. This may be done in a number of ways. Some examples of fruitful collaboration between SLA researchers and classroom teachers are Merrill Swain's work with Maria Kowal (Kowal and Swain 1994, 1997), Catherine Doughty's with Elizabeth Varela (1998), and Joanna White's with CeÂline Goulet (1995). Another way to increase the communication is for researchers to present their work at teachers' conferences and to engage teachers in candid discussion of their ideas. In Quebec, members of the university research community regularly submit a colloquium on `teacher-friendly research' for the meetings of the ESL teachers' association, which attracts mainly primary and secondary school teachers. The audience often includes the teachers in whose classes the research was carried out. In some years we have been disappointed not to have a larger audience since we were in competition with more `practical' workshops. But the evidence suggests that the information presented in these sessions does gradually circulate and become part of teachers' professional knowledge. Since 1985, much SLA has addressed pedagogical concerns, and many young teachers, being trained in university departments where SLA has been given importance, will have SLA as one component of their knowledge base for teaching. This component will shape their expectations about what they can achieve in the classroom. However, it is only when they have tried out some of the pedagogical applications suggested by SLA research that they will understand what it really means for their own teaching context. A study of how grade 6 francophone students in intensive ESL classes in Quebec acquire question forms requires interpretation and re¯ection before it can be seen as relevant to other language learning situations. Indeed, my own students
454 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
constantly ask me about the relevance of our studies of learning in intensive ESL to the learning which can be achieved in the two-hour-a-week ESL classes which are oered to most francophone students in Quebec. How much less might they be immediately interpretable and applicable in an ESL classroom for Hmong-speaking adolescents in Minneapolis? An elementary classroom for Inuktitut-speaking children in Northern Quebec? Students of English as a foreign language in a German Gymnasium? A classroom with 80 students in South Africa? Dierences in both the opportunity and the need to use the language outside of school, dierences in L1 literacy experiences, dierences in L1±L2 language distance, dierences in the organization of the school and classroom, and many other factors contribute to dierences in the kinds of classroom practices which will be eective in dierent contexts. The existence of these dierent realities reinforces the need for more classroom-based research in an even wider range of contexts. As Valdman (1993) argued, there is a great need for replication studies in many areas of research. This need is particularly acute in classroom SLA research. In light of the number of in¯uences on language teaching in dierent contexts, the need for replication studies as well as action research by individual teachers is even greater.7 Nevertheless, it is not the details of the individual studies that can be `applied' but rather the general principles which they re¯ect. SLA research ®ndings do not constitute the only or even the principal source of information to guide teachers in their daily practice of the art and science of second and foreign language teaching. Teachers will make their decisions on the basis of many dierent factors. That has always been true. In 1966, Carroll characterized teachers as, on the one hand, the `standpat traditionalist,' who has found all the right answers and is determined not to let anyone persuade him/her that there is any reason to change and, on the other hand, the `impressionable adventurer,' who is convinced that whatever is new is better than what was there before. Carroll says, `Somewhere between these two extremes I hope we ®nd the majority of teachers, teachers who have convictions about the soundness of their teaching techniques but are open-minded and interested in new ideas, materials, and techniques that stem from research and development, with a readiness to try out these techniques in their classrooms' (1966: 95). No matter how sound the research on which new ideas, materials and techniques are based, pedagogical innovations must be implemented and adapted according to local conditions, the strengths of individual teachers and students, the available resources, the age of the learner, and the time available for teaching. Everybody knows that. Revised version received May 2000
PATSY M. LIGHTBOWN
455
ACKNOWLEDGEMENTS I am deeply grateful to Laura Collins, Howard Nicholas, Nina Spada, Merrill Swain, Joanna White, and three reviewers for Applied Linguistics. They all read earlier versions of the paper and made valuable suggestions for improvements. Claire Kramsch and Martin Bygate provided feedback which was thorough, challenging, and thoughtful. Paul Meara was very generous in helping me co-ordinate the comments of the dierent reviewers. Astrid Liden helped me make the reference list presentable in OUP format. In spite of all this help, I know there are omissions and errors which I never caught. For the omissions, I take refuge in the fact that it would have been impossible to cover all the interesting work on classroom research on second language acquisition. For the errors, I take full responsibility and apologize to anyone whose work is misrepresented by my brief summaries.
NOTES 1 In this paper, for reasons of space and also because of my own research focus, I will concentrate on research which has been carried out with primary and secondary school learners. In addition, and for the same reasons, I will focus on North American research. Happily, Rosamond Mitchell's (2000) anniversary article covers some of the history of language teaching and related research in Britain. 2 It has to be acknowledged that some experienced teachers, especially at the secondary school level, remained sceptical of the dramatic changes in pedagogical practice and made few if any changes in their way of teaching. See also Mitchell and Hooper (1991), Rampton (2000), and Spielmann (1992), for evidence that in many other places as well, CLT has not replaced structure-based teaching. Other teachers took from the new approach what they thought was good and retained what they liked from their previous pedagogy. Nevertheless, new teachers as well as many experienced teachers, especially at the primary school level, embraced CLT with enthusiasm. 3 Allan Forsyth, the person responsible for the development of that innovative comprehension-based approach to ESL learning, did not base his program on Krashen's comprehensible input hypothesis, however. His inspiration for the program design was his observation of his young francophone children's Saturday morning ritual of watching television cartoons in English. He was struck by the fact that they were completely
relaxed and receptive to the language input provided by the short, simple, and entertaining television shows. 4 Of course, many researchers never lost sight of the importance of the ®rst language as one of the signi®cant contributors to the shape of learners' interlanguage, and L1 was an integral part of the original interlanguage hypothesis (Selinker 1972). Even though some argued that L1 had relatively little impact on SLA (e.g., Dulay et al. 1982), others, such as James (1971), Kellerman (1977), Schachter (1974) continued to remind their colleagues that learners' interlanguage might be misinterpreted if L1 in¯uences were not taken into account. 5 As one of the reviewers pointed out, Spada and Lightbown (1999) report that two students in their study `went up 2 stages' between the pre-test and the post-test. There is no reason to believe that students who `went up 2 stages' had skipped the intervening stage. It is at least as likely that we had missed the transition since there was a period of several weeks between the tests. The more general ®nding among the students in the study, however, was that there was no stage change at all. This was attributed to the students following the French-in¯uenced rejection of inversion with nouns which is mentioned under generalization 2. 6 European readers may ®nd this surprising, but French immersion teachers do not always insist on the polite vous when students address them. Furthermore, many
456 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
French Canadian teachers use tu in addressing the whole class: Tu prends ton cahier et tu eÂcris la date. [`You (singular) take your (singular) and you (singular) write the date.'] Thus French immersion students often remain confused about the use of the
dierent forms of `you' even after years of immersion instruction. 7 But see Polio and Gass (1997) for a discussion of some of the reasons such replications are so rare.
REFERENCES Ammar, A. 1996. Is Implicational Generalisation Unidirectional and Applicable in Foreign Contexts? Unpublished M.A. thesis, Concordia University, Montreal. Ausubel, D. 1964. `Adults vs. children in second language learning: Psychological considerations.' Modern Language Journal 48: 420±24. Bardovi-Harlig, K. and Z. DoÈrnyei. 1998. `Do language learners recognize pragmatic violations? Pragmatic vs. grammatical awareness in instructed L2 learning.' TESOL Quarterly 32: 233±59. Bialystok, E. 1994. `Analysis and control in the development of second language proficiency.' Studies in Second Language Acquisition 16: 157±68. Bialystok, E. 1997. `The structure of age: In search of barriers to second language acquisition.' Second Language Research 13: 116±37. Birdsong, D. (ed.) 1998. Second Language Acquisition and the Critical Period Hypothesis. Mahwah, NJ: Lawrence Erlbaum. Bley-Vroman, R. 1983. `The comparative fallacy in interlanguage studies: The case of systematicity.' Language Learning 33: 1±17. Brooks, N. 1964. Language and Language Learning: Theory and Practice, second edition. New York: Harcourt, Brace. Brown, H. D. 1980. Principles of Language Learning and Teaching. Englewood Cliffs, NJ: Prentice Hall. (Second edition published in 1987.) Burstall, C. 1975. `French in the Primary School.' Canadian Modern Language Review 31: 388±402. Carroll, J. B. 1966. `The contributions of psychological theory and educational research to the teaching of foreign languages' in A. Valdman (ed.): Trends in Language Teaching. New York: McGraw-Hill. Chastain, K. 1971. The Development of Modern Language Skills: Theory to Practice. Chicago: Rand McNally.
Chaudron, C. 1977. `A descriptive model of discourse in the corrective treatment of learners' errors.' Language Learning 27: 29± 46. Collier, V. P. 1989. `How long? A synthesis of research on academic achievement in a second language.' TESOL Quarterly 23: 509± 31. Collins, L., R. Halter, P. Lightbown, and N. Spada. 1999. `Time and the distribution of time in second language instruction.' TESOL Quarterly 33: 655±80. Corder, S. P. 1967. `The significance of learners' errors.' International Review of Applied Linguistics 5: 161±69. Crookes, G. 1998. `On the relationship between second and foreign language teachers and research.' TESOL Journal 7(3): 6± 11. Cummins, J. 1984. Bilingualism and Special Education. Clevedon, UK: Multilingual Matters. Cummins, J. 1991. `Interdependence of firstand second-language proficiency in bilingual children' in E. Bialystok (ed.): Language Processing in Bilingual Children. Cambridge: Cambridge University Press. Donato, R. 1994. `Collective scaffolding in second language learning' in J. Lantolf and G. Appel (eds.): Vygotskian Approaches to Second Language Research. Norwood, NJ: Ablex. Doughty, C. and E. Varela. 1998. `Communicative focus on form' in C. Doughty and J. Williams (eds.): Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. Dulay, H., H. Burt, and S. Krashen. 1982. Language Two. New York/Oxford: Oxford University Press. Dussault, B. 1997. Les effets aÁ long terme de l'enseignement intensif de l'anglais, langue seconde. Unpublished M.A. thesis, Universite du QueÂbec aÁ MontreÂal.
PATSY M. LIGHTBOWN
Elley, W. 1989. `Vocabulary acquisition from listening to stories.' Reading Research Quarterly 24: 174±87. Elley, W. 1991. `Acquiring literacy in a second language: The effect of book-based programs.' Language Learning 41: 375±411. Ellis, N. 1995. `Consciousness in second language acquisition: A review of field studies and laboratory experiments.' Language Awareness 4: 123±46. Ellis, N. 1996. `Sequencing in SLA: Phonological memory, chunking and points of order.' Studies in Second Language Acquisition 18: 91±126. Ellis, R. 1989. `Are classroom and naturalistic acquisition the same? A study of the classroom acquisition of German word order rules.' Studies in Second Language Acquisition 11: 305±28. Ellis, R. 1993. `The structural syllabus and second language acquisition.' TESOL Quarterly 27: 91±113. Gass, S. 1982. `From theory to practice' in M. Hines and W. Rutherford (eds.): On TESOL `81. Washington, DC: TESOL. Gass, S. and L. Selinker. (eds.) 1983. Language Transfer in Language Learning. Rowley, MA: Newbury House. Gass, S. and E. Varonis. 1994. `Input, interaction, and second language production.' Studies in Second Language Acquisition 16: 283±302. Genesee, F. 1987. Learning Through Two Languages. New York: Newbury. Green, P. S. and K. Hecht. 1992. `Implicit and explicit grammar: An empirical study.' Applied Linguistics 13: 168±84. Hamilton, R. L. 1994. `Is implicational generalization unidirectional and maximal?' Language Learning 44: 123±57. Han, Y. and R. Ellis. 1998. `Implicit knowledge, explicit knowledge and general language proficiency.' Language Teaching Research 2: 1±23. Harley, B. 1989. `Functional grammar in French immersion: A classroom experiment.' Applied Linguistics 10: 321±51. Harley, B. 1993. `Instructional strategies and SLA in early French immersion.' Studies in Second Language Acquisition 15: 245±59. Harley, B. 1998. `The role of focus-on-form tasks in promoting child L2 acquisition' in C. Doughty and J. Williams (eds.): Focus on
457
Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. Harley, B. and D. Hart. 1997. `Language aptitude and second language proficiency in classroom learners of different starting ages.' Studies in Second Language Acquisition 19: 379±400. Harley, B. and M. L. King. 1989. `Verb lexis in the written compositions of young French L2 learners.' Studies in Second Language Acquisition 11: 415±39. Harley, B. and M. Swain. 1984. `The interlanguage of immersion students and its implications for second language teaching' in A. Davies, C. Criper, and A. P. R. Howatt (eds.): Interlanguage. Edinburgh: Edinburgh University Press. Hatch, E. 1978. `Apply with caution.' Studies in Second Language Acquisition 2: 123±43. Hus, Y. 1997. Central Auditory Processing Disorders in Minority Children. Unpublished PhD dissertation, Concordia University, Montreal. James, C. 1971. `The exculpation of contrastive analysis' in G. Nickel (ed.): Papers in Contrastive Linguistics. Cambridge: Cambridge University Press. Keenan, E. and B. Comrie. 1977. `Noun phrase accessibility and universal grammar.' Linguistic Inquiry 8: 63±99. Kellerman, E. 1977. `Giving learners a break.' Working Papers on Bilingualism 15: 59±92. Kellerman, E. 1985. `If at first you do succeed . . .' in S. Gass and C. Madden (eds.): Input in Second Language Acquisition. Rowley, MA: Newbury. Kellerman, E. and M. Sharwood Smith. (eds.) 1986. Crosslinguistic Influence in Second Language Acquisition. New York: Pergamon. Kelly, L. G. 1969. 25 Centuries of Language Teaching. Rowley, MA: Newbury House. Kowal, M. and M. Swain. 1994. `Using collaborative language prodcution tasks to promote students' language awareness.' Language Awareness 3: 73±93. Kowal, M. and M. Swain. 1997. `From semantic to syntactic processing: How can we promote metalinguistic awareness in the French immersion classroom?' in K. Johnson and M. Swain (eds.): Immersion Education: International Perspectives. New York: Cambridge University Press. Krashen, S. D. 1982. Principles and Practice in
458 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
Second Language Acquisition. Oxford: Pergamon. Krashen, S. D. 1989. `We acquire vocabulary and spelling by reading: Additional evidence for the input hypothesis.' Modern Language Journal 73: 440±64. Krashen, S. D. and T. D. Terrell. 1983. The Natural Approach. Oxford: Pergamon/Alemany. Lado, R. 1964. Language Teaching: A Scientific Approach. New York: McGraw-Hill, Inc. Lambert, W. E. and G. R. Tucker. 1972. Bilingual Education of Children: The St. Lambert Experiment. Rowley, MA: Newbury House. Leeman, J. 2000. `Investigating recasts and L2 development: Negative evidence and enhanced salience.' Paper presented at the American Association for Applied Linguistics, Vancouver, British Colombia, March. Lightbown, P. M. 1985a. `Great expectations: Second language acquisition research and classroom teaching.' Applied Lingusitics 6: 173±89. Lightbown, P. M. 1985b. `Input and acquisition for second-language learners in and out of classrooms.' Applied Linguistics 6: 263±73. Lightbown, P. M. 1991. `What have we here? Some observations on the role of instruction in second language acquisition' in R. Phillipson, E. Kellerman, L. Selinker, M. Sharwood-Smith, and M. Swain (eds.): Foreign/Second Language Pedagogy Research: A Commemorative Volume for Claus Fñrch. Clevedon: Multilingual Matters. Lightbown, P. M. 1992. `Can they do it themselves? A comprehension-based ESL course for young children' in R. CourcheÃne, J. St. John, C. TheÂrien, and J. Glidden (eds.): Comprehension-Based Second Language Teaching. Ottawa: University of Ottawa Press. Lightbown, P. M. 1998. `The importance of timing in focus on form' in C. Doughty and J. Williams (eds.): Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. Lightbown, P. M., R. H. Halter, and J. White. (in preparation) `Comprehension-based learning: The limits of ``Do It Yourself''.' Lightbown, P. M. and N. Spada. 1990. `Focuson-form and corrective feedback in communicative language teaching: Effects on second language learning.' Studies in Second Language Acquisition 12: 429±48.
Lightbown, P. M. and N. Spada. 1991. `EÂtude des effets aÁ long terme de l'apprentissage intensif de l'anglais, langue seconde, au primaire.' Canadian Modern Language Review 48: 90±117. Lightbown, P. M. and N. Spada. 1994. `An innovative program for primary ESL in Quebec.' TESOL Quarterly 28: 563±79. Lightbown, P. M. and N. Spada. (in press) `L1 and L2 in the education of Inuit children in northern Quebec: Students' abilities and teachers' perceptions' in M. Crago (ed.): Language Acquisition, Socialization and Use in Inuit Communities. Littlewood, W. 1984. Foreign and Second Language Learning: Language Acquisition Research and Its Implications for the Classroom. Cambridge: Cambridge University Press. Long, M. 1985. `Input and second language acquisition theory' in S. Gass and C. Madden (eds.): Input in Second Language Acquisition. Rowley, MA: Newbury House. Long, M. 1990. `Maturational constraints on language development.' Studies in Second Language Acquisition 12: 251±86. Long, M. 1991. `Focus on form: A design feature in language teaching methodology' in K. de Bot, R. Ginsberg, and C. Kramsch (eds.): Foreign Language Research in CrossCultural Perspective. Amsterdam: John Benjamins. Long, M. and G. Crookes. 1992. `Three approaches to task-based syllabus design.' TESOL Quarterly 26: 27±56. Long, M. and P. Porter. 1985. `Group work, interlanguage talk, and second language acquisition.' TESOL Quarterly 19: 207±28. Long, M. and P. Robinson. 1998. `Focus on form: Theory, research and practice' in C. Doughty and J. Williams (eds.): Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. Long, M., S. Inagaki, and L. Ortega. 1998. `The role of implicit negative feedback in SLA: Models and recasts in Japanese and Spanish.' The Modern Language Journal 82: 357±71. Lyster, R. 1994. `La neÂgotiation de la forme: StrateÂgie analytique en classe d'immersion.' Canadian Modern Language Review 50: 446±65. Lyster, R. 1998a. `Recasts, repetition, and ambiguity in L2 classroom discourse.' Studies in Second Language Acquisition 20: 51±81.
PATSY M. LIGHTBOWN
Lyster, R. 1998b. `Negotiation of form, recasts, and explicit correction in relation to error types and learner repair in immersion classrooms.' Language Learning 48: 183±218. Lyster, R. and L. Ranta. 1997. `Corrective feedback and learner uptake: Negotiation of form in communicative classrooms.' Studies in Second Language Acquisition 19: 37±66. Mackey, A. and J. Philp. 1998. `Conversational interaction and second language development: recasts, responses and red herrings?' The Modern Language Journal 82: 338±56. Meisel, J., H. Clahsen, and M. Pienemann. 1981. `On determining developmental stages in natural second language acquisition.' Studies in Second Language Acquisition 3: 109± 35. Miller, G. A. 1965. `Some preliminaries to psycholinguistics.' American Psychologist 20: 15±20. Mitchell, R. 2000. `Anniversary Article: Applied linguistics and evidence-based classroom practice: The case of foreign language grammar pedagogy.' Applied Linguistics 21: 281±303. Mitchell, R. and J. V. Hooper. 1991. `Teachers' views of language knowledge' in C. James and P. Garrett (eds.): Language Awareness in the Classroom. Harlow: Longman. Mitchell, R. and C. Martin. 1997. `Rote learning, creativity and `understanding' in classroom foreign language teaching.' Language Teaching Research 1: 1±27. Munby, J. 1978. Communicative Syllabus Design. Cambridge: Cambridge University Press. MunÄoz, C. 1999. `The effects of age on instructed foreign language acquisition' in S. Gonzalez FernaÂndez-Corugedo, R. ValdeoÂn GarcõÂa, D. GarcõÂa Velasco, A. Ojanguren SaÂnchez, A. AntoÂn PeÂrez, and M. Urdiales Shaw (eds.): Essays in English Language Teaching: A Review of the Communicative Approach. Servicio de Publicaciones de la Universidad de Oviedo. Myles, F., J. Hooper, and R. Mitchell. 1998. `Rote or rule? Exploring the role of formulaic language in classroom foreign language learning.' Language Learning 48: 323±63. Myles, F., R. Mitchell, and J. Hooper. 1999. `Interrogative chunks in French L2: A basis for creative construction?' Studies in Second Language Acquisition 21: 49±80. Norris, J. M. and L. Ortega. 2000. `Effective-
459
ness of L2 instruction: A research synthesis and quantitative meta-analysis.' Language Learning 50. Oliver, R. 1995. `Negative feedback in child NS ±NNS conversations.' Studies in Second Language Acquisition 17: 459±81. Pica, T. 1987. `Second-language acquisition, social interaction, and the classroom.' Applied Linguistics 8: 3±21. Pienemann, M. 1985. `Learnability and syllabus construction' in K. Hyltenstam and M. Pienemann (eds.): Modelling and Assessing Second Language Development. Clevedon, Avon: Multilingual Matters. Pienemann, M. 1988. `Determining the influence of instruction on L2 speech processing.' AILA Review 5: 40±72. Pienemann, M. 1999. Language Processing and Second Language Development: Processability Theory. Amsterdam: John Benjamins. Pienemann, M., M. Johnston, and G. Brindley. 1988. `Constructing an acquisitionbased procedure for second language assessment.' Studies in Second Language Acquisition 10: 217±43. Pilgreen, J. and S. Krashen. 1993. `Sustained silent reading with English as a second language high school students: Impact on reading comprehension, reading frequency, and reading enjoyment.' School Library Media Quarterly Fall: 21±23. Polio, C. and S. Gass. 1997. `Replication and reporting: A commentary.' Studies in Second Language Acquisition 19: 499±508. Rampton, B. 2000. `Instructed foreign language rituals in and out of class.' Plenary lecture at the conference of the American Association for Applied Linguistics, Vancouver, British Columbia, March. Ready, D. and M. Wesche. 1992. `An evaluation of the University of Ottawa's sheltered program: Language teaching strategies that work' in R. CourcheÃne, J. St. John, C. TheÂrien, and J. Glidden (eds.): Comprehension-Based Second Language Teaching. Ottawa: University of Ottawa Press. Richards, J. C. 1973. `A non-contrastive approach to error analysis' in J. Oller and J. C. Richards (eds.): Focus on the Learner. Rowley, MA: Newbury. Savignon, S. 1972. Communicative Competence: An Experiment in Foreign-Language Teaching. Philadelphia, PA: The Center for Curriculum Development, INC.
460 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
Schachter, J. 1974. `An error in error analysis.' Language Learning 24: 205±14. Schmidt, R. 1990. `The role of consciousness in second language learning.' Applied Linguistics 11: 17±46. Schmidt, R. 1994. `Deconstructing consciousness in search of useful definitions for applied linguistics' in J. H. Hulstijn and R. Schmidt (eds.): Consciousness in Second Language Learning. AILA Review 11: 11±26. Schwartz, B. D. 1993. `On explicit and negative data effecting and affecting competence and linguistic behavior.' Studies in Second Language Acquisition 15: 147±63. Selinker, L. 1972. `Interlanguage.' International Review of Applied Linguistics 10: 209±31. Sharwood Smith, M. 1993. ` Input enhancement in instructed SLA: Theoretical bases.' Studies in Second Language Acquisition 15: 165± 79. Singleton, D. 1989. Language Acquisition: The Age Factor. Clevedon, UK: Multilingual Matters. Slavoff, G. R. and J. S. Johnson. 1995. `The effects of age on the rate of learning a second language.' Studies in Second Language Acquisition 17: 1±16. Spada, N. 1997. `Form-focussed instruction and second language acquisition: A review of classroom and laboratory research.' Language Teaching 30: 73±87. Spada, N. and P. M. Lightbown. 1989. `Intensive ESL programs in Quebec primary schools.' TESL Canada Journal 7: 11±32. Spada, N. and P. M. Lightbown. 1993. `Instruction and the development of questions in L2 classrooms.' Studies in Second Language Acquisition 15: 205±24. Spada, N. and P. M. Lightbown. 1999. `Instruction, L1 influence and developmental readiness in second language acquisition.' Modern Language Journal 83: 1±22. Spielmann, G. 1992. `A la recherche de l'enseignement communicatif.' The French Review 65: 908±18. Stern, H. H. 1983. Fundamental Concepts of Language Teaching. Oxford: Oxford University Press. Swain, M. 1985. `Communicative competence: Some roles of comprehensible input and comprehensible output in its development' in S. Gass and C. Madden (eds.): Input in Second Language Acquisition. Rowley, MA: Newbury.
Swain, M. 1988. `Manipulating and complementing content teaching to maximize second language learning.' TESL Canada Journal 6: 68±83. Swain, M. 1991. `French immersion and its offshoots: Getting two for one' in B. Freed (ed.): Foreign Language Acquisition: Research and the Classroom. Lexington, MA: Heath. Swain, M. 1995. `Three functions of output in second language learning' in G. Cook and B. Seidlhofer (eds.): Principle and Practice in Applied Linguistics. Oxford: Oxford University Press. Swain, M. 1999. `Language production in SLA: From social to cognitive and back again.' Plenary presented at the Second Language Research Forum, Minneapolis, MN, September. Tarone, E. and M. Swain. 1995. `A sociolinguistic perspective on second language use in immersion classrooms.' The Modern Language Journal 79: 166±78. Tarone, E., M. Swain, and A. Fathman. 1976. `Some limitations to the classroom applications of current second language acquisition research.' TESOL Quarterly 10: 19±32. Terrell, T. D. 1977. `A natural approach to the acquisition and learning of a language.' Modern Language Journal 61: 325±66. Terrell, T. D. 1982. `The Natural Approach to language teaching: An update.' Modern Language Journal 66: 121±32. Trahey, M. and L. White. 1993. `Positive evidence and preemption in the second language classroom.' Studies in Second Language Acquisition 15: 181±204. Truscott, J. 1996. `The case against grammar correction in L2 writing classes.' Language Learning 46: 327±69. Truscott, J. 1999. `What's wrong with oral grammar correction.' Canadian Modern Language Review 55: 437±56. Turnbull, M., S. Lapkin, D. Hart, and M. Swain. 1998. `Time on task and immersion graduates' French proficiency' in S. Lapkin (ed.): French Second Language Education in Canada: Empirical Studies. Toronto, Canada: University of Toronto Press. Valdman, A. 1993. `Replication study (editorial introduction).' Studies in Second Language Acquisition 15: 505. van Lier, L. 1988. The Classroom and the Language Learner. London: Longman VanPatten, B. and T. Cadierno. 1993. `Explicit
PATSY M. LIGHTBOWN
instruction and input processing.' Studies in Second Language Acquisition 15: 225±43. VanPatten, B. and S. Oikkenon. 1996. `Explanation versus structured input in processing instruction.' Studies in Second Language Acquisition 18: 495±510. Weinert, R. 1987. `Processes in classroom second language development: The acquisition of negation in German' in R. Ellis (ed.): Second Language Acquisition in Context. Englewood Cliffs, NJ: Prentice Hall. White, J. 1998. `Getting the learners' attention: A typographical input enhancement study' in C. Doughty and J. Williams (eds.): Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. White, J. 1999. `Focusing the young learner's attention on language form.' Paper presented at Midis du GREDIL, Universite Laval, Sainte-Foy, QueÂbec, October. White, J. and C. Goulet. 1995. `Getting your primary ESL students hooked on books.' SPEAQ-Out 24 (2): 7±13. White, L. 1991. `Adverb placement in second language acquisition: Some effects of positive and negative evidence in the classroom.' Second Language Research 7: 133±61. White, L. and F. Genesee. 1996. `How native is near-native? The issue of ultimate attainment in adult second language acquisition.' Second Language Research 12: 223±65. White, L., N. Spada, P. M. Lightbown, and L. Ranta. 1991. `Input enhancement and L2 question formation.' Applied Linguistics 12: 416±32.
461
Widdowson, H. G. 1978. Teaching Language as Communication. Oxford: Oxford University Press. Wilkins, D. A. 1974. Second-Language Learning and Teaching. London: Edward Arnold. Wilkins, D. A. 1976. Notional Syllabuses. Oxford: Oxford University Press. Wong Fillmore, L. 1991. `Second-language learning in children: A model of language learning in social context' in E. Bialystok (ed.): Language Processing in Bilingual Children. Cambridge: Cambridge University Press. Wray, A. 1999. `Formulaic language in learners and native speakers.' Language Teaching 32: 213±31. Yule, G. and D. Macdonald. 1990. `Resolving referential conflicts in L2 interaction: The effect of proficiency and interactive role.' Language Learning 40: 557±99. Zimmerman, C. B. 1997. `Do reading and interactive vocabulary instruction make a difference? An empirical study.' TESOL Quarterly 31: 121±40. Zobl, H. 1979. `Nominal and pronominal interrogation in the speech of adult francophone ESL learners: Some insights into the workings of transfer.' SPEAQ Journal 3: 69± 93. Zobl, H. 1980a. `Developmental and transfer errors: Their common bases and (possibly) differential effects on subsequent learning.' TESOL Quarterly 14: 469±479. Zobl, H. 1980b. `The formal and developmental selectivity of L1 influence on L2 acquisition.' Language Learning 30: 43±57.
APPENDIX Some of the books, published since 1985, which are devoted to classroom research on SLA or the impact of SLA research on classroom learning. Allwright, D. and K. M. Bailey. 1991. Focus on the Language Classroom: An Introduction to Classroom Research for Language Teachers. Cambridge: Cambridge University Press. Brumfit, C. and R. Mitchell. 1990. Research in the Language Classroom. ELT Documents (133). Modern English Publications in Association with the British Council. Chaudron, C. 1988. Second Language Classrooms: Research on Teaching and Learning. Cambridge: Cambridge University Press.
Cook, V. 1996. Second Language Learning and Language Teaching 2nd edn. London: Arnold. Doughty, C. and J. Williams (eds.) 1998. Focus on Form in Classroom Second Language Acquisition. Cambridge: Cambridge University Press. Eckman, F., D. Highland, P. W. Lee, J. Mileham, and R. R. Weber. (eds.) 1995. Second Language Acquisition Theory and Pedagogy. Mahwah, NJ: Lawrence Erlbaum. Ellis, R. 1990. Instructed Second Language Acquisition. Oxford: Basil Blackwell.
462 CLASSROOM SLA RESEARCH AND SECOND LANGUAGE TEACHING
Ellis, R. 1991. Second Language Acquisition and Language Pedagogy. Clevedon, Avon: Multilingual Matters. Ellis, R. 1997. SLA Research and Language Teaching. Oxford: Oxford University Press. Freed, B. (ed.) 1991. Foreign Language Acquisition: Research and the Classroom. Lexington, MA: Heath. Johnson, D. 1992. Approaches to Research in Second Language Learning. New York: Longman. Johnson, R. K. and M. Swain. 1997. Immersion Education: International Perspectives. Cambridge: Cambridge University Press. Lantolf, J. and G. Appel. (eds.) 1994. Vygotskian Approaches to Second Language Research. Norwood, NJ: Ablex. Lightbown, P. M. and N. Spada. 1999. How Languages Are Learned 2nd edn. Oxford: Oxford University Press.
Nunan, D. 1989. Understanding Language Classrooms. New York: Prentice Hall. Nunan, D. 1992. Research Methods in Language Learning. Cambridge: Cambridge University Press. Phillipson, R., E. Kellerman, L. Selinker, M. Sharwood Smith, and M. Swain. (eds.) 1991. Foreign/Second Language Pedagogy Research. Clevedon, UK: Multilingual Matters. Schachter, J. and S. Gass. 1996. Second Language Classroom Research. Mahwah, NJ: Lawrence Erlbaum. Spada, N. and M. FroÈhlich. 1995. COLT: Communicative Orientation of Language Teaching Observation Scheme: Coding Conventions and Applications. Sydney, Australia: National Centre for English Language Teaching and Research. Van Lier, L. 1988. The Classroom and the Language Learner. London: Longman.
Applied Linguistics 21/4: 463±489
# Oxford University Press 2000
Formulaic Sequences in Second Language Teaching: Principle and Practice ALISON WRAY Cardi University, Cardi, UK One important component of successful language learning is the mastery of idiomatic forms of expression, including idioms, collocations, and sentence frames (collectively referred to here as formulaic sequences). Three attempts to foreground formulaic sequences in teaching syllabuses are those of Willis (1990), Nattinger and DeCarrico (1992), and Lewis (1993). All three ®nd themselves confronting the question of how the teaching of multi-word strings relates to the learner's accumulation of grammatical and lexical knowledge, and despite their dierent viewpoints and priorities, all conclude that larger units can, and should, be perceived by the learner and teacher in terms of their component parts. Yet research into the nature of formulaic sequences indicates that their form often precludes, and their function speci®cally circumvents, such internal inspection, for their value resides in the bypassing of the analytical processes which encode and decode strings. Thus, Willis, Nattinger and DeCarrico, and Lewis are all pursuing native-like linguistic usage by promoting entirely unnative-like processing behaviour. This non-alignment is only tractable if the classroom teaching of languages is fully acknowledged as arti®cial, even when the methods used appear `naturalistic'.
INTRODUCTION In recent years there has been increased interest in targeting formulaic language in second language teaching. In its narrowest sense, formulaicity has always been a useful entrance point for the learner, with the `phrase-book' approach providing a few pre-learned utterances for asking the way to the station or ordering a cup of coee. However, there is much more to it than this. Gaining full command of a new language requires the learner to become sensitive to the native speakers' preferences for certain sequences of words over others that might appear just as possible. From the bizarre idiom, through the customary collocation, to the turns of phrase that have no other apparent linguistic merit than that `we just say it that way', the subtleties of a language may ¯oor even the pro®cient non-native, not so much because of a non-alignment between interlanguage and target language forms, as because the learner lacks the necessary sensitivity and experience that will lead him or her unerringly away from all the grammatical ways of expressing a particular idea except the most idiomatic. For example, Pawley and Syder (1983: 195), who explore this learner problem in some detail, contrast the expression I'm
464 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
so glad you could bring Harry with several grammatical equivalents which are simply never said, including, That Harry could be brought by you makes me so glad and That you could bring Harry gladdens me so. In this paper, I want to examine the assumptions behind three attempts to introduce formulaic language into second language teaching. In the remainder of this section, I contextualize the discussion by brie¯y outlining the nature of formulaic language as a phenomenon. The following section compares and evaluates the assumptions underlying the three syllabuses: Willis (1990), Nattinger and DeCarrico (1992), and Lewis (1993). In the third section, I explore the nature of formulaic language in more detail, with reference to these assumptions, before, in the fourth section, confronting the three approaches with a fundamental incongruity, and considering whether it can be resolved.
A basic characterization of formulaic language Many attempts have been made to categorize formulaic language (e.g. Aijmer 1996, Becker 1975, Bolinger 1976, Coulmas 1979, 1994, Hatch et al. 1979, Howarth 1998a and b, Hudson 1998, Lattey 1986, Moon 1992, 1998a and b, Nattinger and DeCarrico 1992, Van Lancker 1987, Yorio 1980). Most recognize, explicitly or implicitly, the need to dierentiate between form and function, though it proves repeatedly dicult to separate the two. A full appreciation of what formulaic language is requires us to recognize that we are not dealing with a single phenomenon, but rather with a set of more and less closely related ones, across dierent data types (see Wray 1999, Wray and Perkins 2000), including the output of ®rst language learners (e.g. Bates et al. 1988, Nelson 1981, Peters 1977, 1983, Plunkett 1993), second language learners (e.g. De Cock et al. 1998, Ellis 1984, Granger 1998, Howarth 1998a and b, Myles et al. 1998, 1999; Raupach 1984; Weinert 1995; Wong Fillmore 1976), adult natives (e.g. Becker 1975, Bolinger 1976, Coulmas 1979, Cowie 1988, 1994a and b) and the linguistically disabled (e.g. Code 1987 1997, Hughlings Jackson 1874/1958, Perkins 1994, 1999, Van Lancker 1987, Van Lancker and Kempler 1987). The formulaic language observed in such studies has been subject to independent labelling, but with some measure of crossborrowing between ®elds. The result is a huge set of descriptional and de®nitional terms, including those in Table 1, which can give the unwary a false impression that what has been found in one type of speaker is the same as, or de®nitely dierent from, what has been found in another (see Wray 1999, in press, Wray and Perkins 2000). It certainly is not safe to claim, as Weinert (1995: 182) does, that `while labels vary, it seems that researchers have very much the same phenomenon in mind'. It might seem, in the light of Table 1, that the last thing we need is yet another term, yet without one there is a practical terminological problem in evaluating what has gone before. Every label has its own history and implications, and neutral and inclusive reference is impossible without a clear
ALISON WRAY 465
Table 1: Terms used to describe aspects of formulaicity in the literature amalgams
gambits
automatic chunks clicheÂs co-ordinate constructions collocations composites conventionalized forms
gestalt holistic holophrases idiomatic idioms irregular lexical(ized) phrases
F[ixed] E[xpressions] including I[dioms] ®xed expressions
lexicalized sentence stems multiword units
formulaic language formulaic speech formulas/formulae fossilized forms
non-compositional non-computational non-productive non-propositional
frozen metaphors frozen phrases
petri®cations praxons
preassembled speech prefabricated routines and patterns ready-made expressions ready-made utterances recurring utterances rote routine formulae schemata semi-preconstructed phrases that constitute single choices sentence builders stable and familiar expressions with specialized subsenses stereotyped phrases stereotypes stock utterances synthetic unanalysed chunks of speech
and dedicated one. In what follows, therefore, I shall use the term formulaic sequence to encompass the wide range of phenomena variously labelled in the published literature. My de®nition of the formulaic sequence is as follows: a sequence, continuous or discontinuous, of words or other meaning elements, which is, or appears to be, prefabricated: that is, stored and retrieved whole from memory at the time of use, rather than being subject to generation or analysis by the language grammar.1
In eect, this means that the words in a formulaic sequence are `glued together' and stored as a single `big word' (Ellis 1996: 111). In many cases, sequences will have been learned this way in the ®rst place. However, some appear to be constructed out of their individual constituents using the
466 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
grammar, and only then turned into a formulaic sequence, in a process called fusion (Peters 1983). It has been acknowledged for many years that idioms, at the very least, must be viewed as subject to this `big words' de®nition (e.g. Chafe 1968; Weinreich 1969; see also discussion in Hudson 1998: 29.). In particular we may note the ones which are semantically opaque, such as beat about the bush, or syntactically irregular, such as by and large. In both cases, it is implausible that they are generated by rule out of their lexical components. However, these irregular idioms are a small group, relative to the larger phenomenon. In order to encompass the whole range, it is necessary to allow for the possibility that word sequences may be formulaic even though they do not need to be, that is, even though they are semantically transparent and syntactically regular. Taking this stance means that sequences like It was lovely to see you or There are three things to consider . . . Firstly . . . Secondly . . . Thirdly . . . can be viewed as potentially formulaic, even though they could be fully generated from their smallest components by the language grammar at the time of use (Sinclair 1991). Allowing for formulaic sequences to be regular in their form and meaning has a number of signi®cant corollaries, three of which we need to explore here. The ®rst is that, from dealing with a peripheral set of slightly awkward idioms, we shift to the possibility that any quantity of our language could be formulaic. Corpus research (e.g. Butler 1997; Eeg-Olofsson and Altenberg 1994; Moon 1998a and b; Stubbs 1995, 1997) has made a number of estimates of the proportion of formulaic material in normal language, stretching as high as 80 per cent (Altenberg 1998). The ®gures vary, inevitably, according to precisely what is being counted and the nature of the texts examined, including the proportion of written to spoken material, for although speech and writing both feature formulaic sequences, the forms and distributions are dierent (Butler 1998: 28; Moon 1998b: 72f).2 The second corollary is an extension of the ®rst. If language of regular composition can be formulaic, then formulaic and newly generated language will look identical, so how are we to tell them apart? Corpus-based estimates are built upon the assumption that formulaic sequences can be identi®ed by virtue of their being more frequent than other word strings, indeed, that frequency is a central de®nitional criterion. Although this is a not unreasonable starting place, there are some diculties with it. One is that there are undoubtedly some formulaic sequences that are widely accepted as such by native speakers but which are actually not very frequent in normal discourse. These include those associated with a speci®c high pro®le event or cultural phenomenon, e.g. The King is dead, long live the King; All for one and one for all. Moon (1998a) records that when she searched the 18 million word Hector corpus for occurrences of the 6,700 expressions which are listed as phrases in the Collins COBUILD English Language Dictionary, she found 70 per cent to occur with a frequency of less than one in a million, and 40 per cent not at all (p. 82). Another diculty with using frequency as a means of spotting formulaicity is that it forces us to
ALISON WRAY 467
assume that any sequence of words that is repeated a few times is formulaic, that is, that we will not generate the same sentence from scratch very often without then keeping a copy whole for later use. There may be some reasons for favouring this idea (see `The functions of formulaic sequences' below), but it certainly is not safe to assume it. The diculty, in short, is that formulaic and non-formulaic language may sometimes look identical, and frequency counts may not be a reliable means of dierentiating them. Indeed, the vexed question of how to identify formulaicity in a consistent and principled way is far from being solved. Apart from frequency, another possible indicator is phonological form. It has been noted that children, certainly, enunciate less clearly those chunks of language which they have learned whole and for which they have not yet acquired control of the constituent components (e.g. Peters 1983, Plunkett 1993), and it could be argued that the phrases in which consonant weakening and vowel reduction are found in the continuous speech of adult natives, e.g. [
e'sõÄzt@bI] (there seems to be); [fµTek@bI] (if there could be) (Brown 1990: 83) are in actual fact formulaic sequences. If so, a formulaic sequence might conceivably be identi®ed as the stretch of language undergoing such phonological eects, bordered by sounds which do not. Another useful insight into identi®cation, though with limited application, comes from Backus (1999), who suggests that when bilinguals code-switch, they do so at the boundaries of formulaic sequences. Most of the other candidate identi®cational criteria (see, for example, Weinert 1995: 182f) focus on the immature or interlanguage forms of learners, where fossilized errors or the correct use of a construction otherwise not within the scope of the speaker's grammatical competence can indicate that the sequence has not been created from scratch at the time. In actual fact, it is probable that a satisfactory means of identi®cation will entail more than one diagnostic, and this makes it particularly important to understand as fully as possible what formulaic sequences are, so that there is no danger of circularity between the de®nitional and the identi®cational criteria. The third corollary of the proposal that semantically transparent and grammatically regular strings can be formulaic is that speakers and hearers must be engaging in strategy choices during processing, because all but a few sequences (the idioms at one end and what Altenberg's ®gure implies to be 20 per cent of non-formulaic strings at the other) could be processed either way. This means that a satisfactory account of the way formulaicity works in language must engage with both psycholinguistic and general linguistic theory, to accommodate two interacting systems of storage, retrieval and generation. We shall return to this later. As for what formulaic sequences are for, we shall consider this in the third section. We turn now to an examination of three recent approaches to the teaching of formulaic sequences.
468 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
THE TEACHING OF FORMULAIC SEQUENCES There is general consensus that formulaic sequences are extremely dicult for the L2 learner to master (Moon 1992, Scarcella 1979, Yorio 1980, 1989). Kuiper and Tan Gek Lin (1989) see an inherent link between native-like competence and native culture, to the extent that they doubt, given the complexity of the cultural information which is encoded in formulae, that anyone can become truly bicultural after early childhood and therefore that anyone can become a native speaker of a second language after this time even if they sound as though they are (Kuiper and Tan Gek Lin 1989: 304).
Knowing which subset of grammatically possible utterances is actually commonly used by native speakers is an immense problem for even the most pro®cient of non-natives, who are unable to separate out and avoid the grammatical but non-idiomatic sequences (Pawley and Syder 1983). If formulaic sequences are so dicult to learn, then unless we understand why, we are unlikely to hit on a successful way of teaching them. One possible cause is the poverty of the learner experience. Irujo (1986) points out that formulaic sequences are `frequently omitted in the speech addressed to second-language learners' and although they are common in television and movies, `input without interaction is not sucient for language acquisition' (pp.236±7). Furthermore, they are often not taught very well (Granger 1998, Irujo 1986: 237) and it is easy for the wrong ones to be taught (Williams 1988: 51). The learner may appear non-native by virtue of having either too few or too many (Granger 1998). It seems dicult to match in the classroom the `real world' experience of language, whereby it might be possible for observation and imitation to lead the learner to prefer those sequences which are the usual forms in a given speech community (e.g. Willis 1990: 63f).
Approaches to teaching formulaic sequences Despite the diculties, three recent attempts have been made to introduce formulaic sequences into the teaching programme in a principled way. Each of these approaches has its own priorities and sees rather dierently the role of formulaicity in language knowledge. Nattinger and DeCarrico (1992: 117) are mostly interested in the interactional functions associated with individual examples of common formulaic sequences. They focus on their usefulness in teaching conversation, recommending the following steps: . pattern practice drills using ®xed routines, to develop con®dence and ¯uency; . controlled variation using substitution drills to demonstrate that `the chunks learnt previously were not invariable routines, but were instead patterns with open slots'; . increased variation `allowing them to analyze the patterns further'.
ALISON WRAY 469
Lewis (1993) embraces the formulaic sequence by downgrading the signi®cance of the single word as a unit, preferring the broader term lexical item, which encompasses formulaic word strings too. In his approach, it is `lexical phrasesÐa particular kind of lexical itemÐ. . . [that] provide the basis for a lexically . . . driven syllabus' (p. 100). He selects lexical phrases on the basis of their `archetypicality', and aims to provide: . `a large vocabulary, even if [low level students] are initially unable to grammaticalize it'; . `pragmatically useful lexical items, particularly institutionalized utterances'; . `a balance . . . between (relatively rare) words carrying considerable meaning, and (relatively wide and frequent) patterns with low meaning content' (Lewis 1993: 106±7).
Willis (1990) is less interested in word strings per se than in the ways in which certain words ®gure within them. He favours `procedures which make [the] patterns salient' (personal communication). He believes that `we need to help students to notice patternings and to speculate about them' (ibid). His approach introduces formulaic sequences incidentally, as part of the body of data used to demonstrate words in their customary usage. The rationale is clear and sensible: The commonest patterns in English occur again and again with the commonest words in English. If we are to provide learners with language experience which oers exposure to the most useful patterns of the language, we might as well begin by researching the most useful words in the language (Willis 1990: 38).
He pivots his material on key words, selected for their frequency (a procedure also favoured by Sinclair and Renouf 1988). For example, he shows how way, the third most common noun in English (after time and people) (p. 28) appears in characteristic ®xed phrases such as by the way, by way of (p. 30) and frames such as the best way to . . . is to . . . ; One way of . . . -ing . . . is by . . . -ing (p. 38). Thus, Willis' attention is focused, much more than that of the others, on part®xed, part-variable strings, including the lexicalized sentence stems of Pawley and Syder (1983) described later.
Underlying assumptions Formulaic sequences as a means of accessing the grammar and lexicon While Willis' approach deliberately associates numerous common wordstrings according to their shared possession of a word, Lewis and Nattinger and DeCarrico identify the word as the key to grammar. Lewis focuses on the value of `word grammar (collocation and cognates)' as opposed to `structure' as the teaching currency (1993: 3) and favours the view of `grammar as a receptive skill, involving the perception of similarity and dierence' (p. vii), as
470 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
opposed to traditional rule-based approaches. Nattinger and DeCarrico point to Oller's notion of a `grammar of expectancy' (p. 34), whereby knowledge of a language derives in part from being able to gauge accurately the measure of variation possible in any given sequence, something achieved by experience, rather than the application of grammatical rules. Indeed, they perceive no solid distinction between the domain of the lexical item and that of the grammatical rule in determining acceptable sequences (pp. 22f). Most signi®cantly, Nattinger and DeCarrico speci®cally claim that adult second language learners can be supposed to use formulaic sequences as input for their analysis of the language, out of which they will derive grammatical and morphological rules (pp. 27). In similar vein, Lewis states that: `grammar will, to some extent at least, be acquired through generalizing, and learning the restrictions on the generalization from these sentences' (1993: 100). Willis (1999) suggests that: `we need to encourage learners to analyse the language they have experienced in such a way as to facilitate development and to inculcate productive approaches to learning' (pp. 117±18). Their collective position is supported by Bolander (1989), who proposes that `when the number of prefabs stored in memory is large enough, syntactic rules are derived as help for the memory to economize and rationalize processing' (p. 85). In other words, pressure on memory forces a reorganization of the language data into sub-units suitable for fast and ¯exible manipulation. This is reminiscent of accounts of the cause of the vocabulary spurt in L1 acquisition (e.g. Plunkett 1993) and, more particularly, of the kickin of the L1 grammar (Locke 1993; Wray and Perkins 2000). In contrast, neither Scarcella (1979), Schmidt (1983), Yorio (1989) nor Granger (1998) considers formulaic sequences to be useful contributors to the underlying grammar knowledge in adults. Scarcella believes that it is a mistake for the learner to `approach routines as though they were analysable rather than unanalysable wholes' (p. 83) because, in encouraging the production of novel forms by analogy, it will lead to many erroneous ones. Granger (1998) states that: there does not seem to be a direct line from prefabs to creative language. . . . It would thus be a foolhardy gamble to believe that it is enough to expose L2 learners to prefabs and the grammar will take care of itself (pp. 157±8).
Care must be taken, of course, to dierentiate between the idioms, which cannot be usefully analysed, and to which Scarcella (1979) for one is probably referring, and the much larger body of `regular' sequences which, even if normally processed holistically, could oer a learner insights into the grammatical system when analysed. The diculty for the learner is dierentiating the regular from the irregular without already knowing the grammatical patterns of the language. To summarize, there is disagreement regarding the ability of taught learners to make generalizations about the grammar from formulaic input. The teaching syllabuses tend to assume it,
ALISON WRAY 471
whereas experimental, corpus and observational research seem to cast doubt on it (pace Myles et al. 1998, 1999, discussed below).
Child and adult learners Part of the justi®cation for believing that adults can use formulaic sequences as raw material for analysing and learning the language grammar is the assumption that adults and children use formulaic sequences in the same way. However, the evidence to date suggests that the situation is more complicated than this. It is certainly the case that formulaic sequences ®gure in ®rst language acquisition, even if only in the sense of early imitations of imperfectly understood, and non-segmented, adult utterances. In some children they appear to be a major acquisitional strategy that both delivers to the child the capability of achieving lengthy interactional turns well beyond the scope of its current grammatical knowledge and provides coherent linguistic material, with an associated situational interpretation, which can be recalled at will and subjected to segmentation, identifying useful recombining components (Bretherton et al. 1983, Hickey 1993, Lieven et al. 1992, Nelson 1981, Peters 1977, 1983, Plunkett 1993). Formulaic sequences have also been found in young children learning a second language (e.g. Bates and MacWhinney 1987, Hakuta 1974, Huang and Hatch 1978, Peters 1983, Wong Fillmore 1976, see Wray 1999, in press for reviews), where, again, there is indeed evidence of their being broken down into their constituent parts and used creatively. The question is, to what extent are such observations helpful in understanding the language behaviour of taught teenagers and adults? The literature abounds with considerations of the ways in which the young child's L2 learning is qualitatively dierent from that of the adult. Brain plasticity, stage of conceptual development, level of physical dependency, stage of literacy, sense of social identity, quality and quantity of input, aective factors, and so on, have all been cited as possible determiners of a dierent starting place, route, target and ultimate attainment level for the child and the adult learner (e.g. Ellis 1994). Yet Nattinger and DeCarrico (1992) have no hesitation in equating child and adult data: there is no reason to think that adults would go about the task completely dierently [from children]. In important ways, the language learning situation is the same for adults as for children, and makes it likely that an adult learner would also ®nd prefabricated language an ecient way to begin to acquire a new language system (p. 27).
Many of the published commentaries on formulaic sequences in L2 (e.g. Granger 1998, Krashen and Scarcella 1978, Nattinger and DeCarrico 1992, Weinert 1995) follow this trend, and amalgamate evidence from learners of dierent ages and types. Weinert (1995), for example, juxtaposes data from child, teenage, and adult learners, and the latter in both naturalistic and
472 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
classroom settings, in her discussions of the functions of formulaic language (pp. 186). The eect, albeit unintentional, is to suggest stronger trends in the data than there actually are, and to miss some striking dierences in the patterns of form and function between dierent types of learner (Wray 1999). Once the dierent data types are separated out, it becomes clear that there is very little evidence that adult learners naturally extrapolate grammatical or lexical information from larger strings. If we look ®rst at naturalistic learners, it seems that, indeed, many use little formulaic language at all. Yorio's (1989) survey, for example, `[did] not appear to ®nd extensive use of prefabricated language in untutored adult learners' (p. 57). Where formulaic sequences do play more of a role in the development of communicative competence, this usage does not necessarily lead to grammatical accuracy (e.g. Schmidt 1983). Indeed, formulaic sequences may be used by some adult learners as a means of actually avoiding engaging with language learning. Rehbein's (1987) Turkish Gastarbeiter in Germany used formulaic sequences to achieve the most minimal level of communication that they could survive on, accompanied by a `self-imposed reduction of their own system of needs' (p. 245).3 Hinnenkamp (1980), similarly, found that Gastarbeiter resisted learning the language, preferring a partly formulaic `pidgin'. Overall, adult naturalistic learners seem able to draw on a range of strategies for coping with the diculties of communication, only one of which is engagement with language learning via formulaic sequences. Research on classroom-taught learners does indicate more consistency in the use of formulaic sequences, whether as a means of sustaining basic interaction (e.g. Bolander 1989, Bygate 1988, DeCock et al. 1998, Myles et al. 1998), or in pursuit of the learner's desire to produce correct forms (Biskup 1992, Weinert 1995: 193). However, the question of whether classroom learners are able to break down and analyse formulaic sequences without explicit instruction has hardly been investigated. One notable exception is a recent longitudinal study of sixteen 11 and 12 year-old learners of French in a British school (Myles et al. 1998, 1999). The chunk comment t'appelles-tu?, to take one example, was found to be used for third as well as second person reference at ®rst, but after some time the two pronouns in the sequence were, independently, replaced with the correct forms. Although the evidence is promising, it is, unfortunately, unable to throw much light on our current question. This is because the subjects were also given explicit tuition on the very forms they ®nally preferred (e.g. comment s'appelle-t-elle?), so it is dicult to tell whether any segmentation that occurred resulted in or was itself the result of an emerging awareness of the individual meaning components of the formulaic sequence. In sum, despite the claims of Nattinger and DeCarrico, and the assumptions of others, there is only limited support for the idea that evidence from child L2 learners is a good predictor of how adults will handle formulaic material.
ALISON WRAY 473
Side-stepping the grammar The above discussion indicates that there is, as yet, little reliable evidence for adult classroom learners being able to infer grammatical structure from holistically-learned sequences without explicit guidance. Yet Willis (1990), Nattinger and DeCarrico (1992), and Lewis (1993) all place their focus on the way in which words in association may oer natural access to grammatical knowledge. This emphasis re¯ects a more general trend in language research, to loosen the grip of syntax-heavy theory and embrace some of the more messy and elusive patterns of language use that have surfaced as a result of corpus research such as that of the COBUILD project. However, syllabus writers are constrained by the need to ful®l the standard requirements and expectations of language teaching programmes, including an account of the mechanisms by which the learners will develop grammatical knowledge. The rather optimistic answer, that it will occur by natural inferencing (Lewis 1993: vi, Willis 1990: vii), side-steps the need to supply explicit analytic material alongside the holistic (as, for example, Yorio 1980: 434 recommends), while acknowledging that it must be accounted for. This is an olive-branch to grammar-championing conservatives and also serves to distance them from the discredited audio-lingual method (see for example Nattinger and DeCarrico 1992: 116, Richards and Rodgers 1986: 58). But why should formulaic sequences not be reliable in supplying the information about constituents that the learner needs in order to develop a full linguistic system? In order to answer this question, we need to examine in more detail the nature of formulaic sequences and, in particular, their function.
THE FUNCTIONS OF FORMULAIC SEQUENCES Two main explanations for formulaicity in language are found in the literature. These are independently derived and have not previously been accommodated within a single explanatory model. We shall brie¯y examine each, and how they may relate to each other.
Saving eort in processing The ®rst function identi®ed for formulaic sequences is that they save precious processing resources: [they] give us ready-made frameworks on which to hang the expression of our ideas, so that we do not have to go through the labor of generating an utterance all the way out from S every time we want to say anything (Becker, 1975: 17).
Such savings in processing seem to be valuable, particularly during demanding concurrent tasks (Wray 1990, 1992). Not only continuous stretches of language fall into this category. As Pawley and Syder (1983)
474 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
show, whole paradigms of potential utterances can be based on a single lexicalized sentence stem. One of their examples is NP be-TENSE sorry to keepTENSE you waiting (p. 210), which gives us I'm sorry to keep you waiting; Mr Smith will be sorry to have kept you waiting, and so on. In such discontinuous formulaic sequences, there are gaps for lexical items, which may be entirely open to any semantically plausible member of the word class, or may be subject to collocational restrictions, as with pay attention and take care but not *take attention or *pay care. Even sequences with no open class slots nevertheless mostly accept morphological variation, ®tting tense, person and number to the context (e.g. I've done really well, you'll do really well, they did really well, etc.). Although there must be some analytic processing involved in slotting words or morphological forms into an established frame, there is, the argument goes, less eort involved in this than in creating the whole construction from scratch. Table 2 represents Wray and Perkins' (2000) subcategorization of the processing functions of formulaic sequences into three types. The ®rst is the short-cutting process. It helps explain why we prefer certain collocations, and why an individual's speech often features personal characteristic phrases. It also accounts for how the rehearsal of a speech or lecture can lead to the establishment of an informal script, and how we may, by the third or fourth telling of a story, be virtually reciting our previous account word for word. Recent research on brain activity in novel and familiar linguistic tasks con®rms that a practised routine does not just speedily access the original processing route, but actually by-passes it (McCrone 1999, Raichle 1998). The second category in Table 2 relates to (a) sequences whose very bulk seems to oer advantages in the construction of discourse, perhaps by providing rhetorical balance, or by pacing the appearance of novel material, and (b) the ®llers, turn-holders and discourse markers which enable us to carve out a temporal space for the construction of our novel message, by stalling for time and registering a claim to be heard. The third category is slightly dierent. We reduce strain on our memory by holding information inside a formulaic sequence, so that it will be easy to recall later.
Achieving interactional functions The second explanation of formulaicity oered in the literature relates not to processing but to the act of communication. A number of taxonomies of formulaic sequences (or subsets of them), most notably those of Nattinger and DeCarrico (1992) and Aijmer (1996), are based on the socio-interactional function that some of them have, for greeting, thanking, apologizing, and so on. Many such functions seem to rely on the use of agreed forms to a greater or lesser extent. Table 3, based on one in Wray and Perkins (2000), identi®es three types of socio-interactional function for formulaic sequences. All relate to aspects of how we want others to treat or view us. The ®rst re¯ects our need to change our physical world through the actions of another person, and
Table 2: The functions of formulaic sequences in reducing processing eort Function
Eects
Types
Examples
Processing shortcuts
Increased production speed and ¯uency
· Best foot forward; I have known _ for _ years in my capacity as _ · Personal computer; bullet point; the current economic climate
Time-buyers
a) Vehicles for ¯uency, rhythm & emphasis b) Planning time without losing the turn
· standard phrases (with or without gaps) · standard referential labels with agreed meanings · standard phrases with simple meanings · ®llers · turn-holders · discourse shape markers · repetitions of preceding input
Gaining and retaining access to information otherwise unlikely to be remembered
· mnemonics · lengthy texts one is required to learn · rehearsal
ALISON WRAY 475
Manipulation of information
· Make a decision; at the end of the day (in the sense of `really'); one way and another · If the truth be told; if you want my opinion; if you like · And another thing; and let me just say . . . · There are three points I want to make. Firstly . . . Secondly . . . Thirdly/Lastly · (A: What's the capital of Peru?) B: What's the capital of Peru? (Lima isn't it?) · Thirty days hath September . . . ; Richard of York gave battle in vain · Shall I compare thee to a summer's day? · Rehearsing a telephone number while looking for a pen
Function
Eects
Types
Examples
Manipulation of others
satisfying physical, emotional and cognitive needs
Asserting separate identity
a) being taken seriously
· · · · ·
· · · · ·
Asserting group identity
commands requests politeness markers bargains, etc. story-telling skills
· turn claimers etc.
·
b) separating from the crowd
· personal turns of phrase
·
a) overall membership
· `in' phrases
·
· group chants · institutionalized forms of words, etc. · ritual · proverbs
· ·
· threats · quotation
· ·
· forms of address · hedges etc.
· ·
b) place in hierarchy (arming and adjusting)
· ·
keep o the grass; hand it over could you repeat that please? I wonder if you'd mind . . . I'll give you __ for it You're never going to believe this, but . . . Yes, but the thing is . . . ; Thank you very much (in response to invitation to speak); The ®rst thing that you have to realize, of course, in addressing this issue is . . . I wanna tell you a story (Max Bygraves); You know what I mean, Harry (Frank Bruno) Praise the Lord!; as the actress said to the bishop We are the champions Happy birthday; dearly beloved, we are gathered here today . . . Our Father, which art in Heaven . . . Look before you leap; A stitch in time (saves nine) I wouldn't do that if I were you ``I don't want to belong to any club that will have me as a member'' (Groucho Marx) Your Highness Well I'm not sure (as a polite denial or refusal)
476 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
Table 3: The functions of formulaic sequences in social interaction
ALISON WRAY 477
includes commands, requests, bargains, and the frames which characterize politeness in these contexts. The other two functions relate to the personal identity of the speaker and the intention to manipulate the hearer's perception of him/her. Formulaic sequences can be used to assert both our individual and our group identity. They enable us to make statements about our sameness and dierentness, and to jostle for position within the hierarchy. There is a striking similarity between the socio-interactional functions described in Table 3 and the functions achieved through the noise-gesture communication of chimps (Reiss 1989).4
How the functions relate Although, between them, the processing advantages and the interactional functions can oer a rationale for most, if not all, formulaic sequences, the relationship between these two motivations is not immediately clear. Are they any more than coincidental associates? Certainly, they seem to dier in kind too much to be truly complementary. It is dicult to conceive of prefabricated sequences that would not confer processing advantages, and that seems to imply that these must constitute a universal set, of which sequences with a socio-interactional function can only be a subset. However, Wray and Perkins (2000) propose one way of accommodating the two functions within a single model. When an individual chooses a prefabricated stretch of language in order to reduce the pressure on processing, the aim is to be ¯uent and to succeed in producing the entire message without interruption, or to ensure that information is reliably to hand when needed. In other words, the speaker selects a formulaic sequence in the interests of ecient production. In contrast, when the speaker selects a formulaic sequence for socio-interactional purposes, what is paramount is the eect of the words on the hearer. The success of the utterance is not measured by whether it is delivered clearly and ¯uently, nor even whether its internal composition is extricated and fully appreciated, but rather by whether the hearer reacts appropriately, be that in providing a requested item, backing down in the face of a threat, or registering the speaker as an in- (or out-) group member. In other words, the selection of a formulaic sequence in the socio-interactional context aims to achieve the maximum chance of ecient comprehension. While the formulaic sequences represented by the categories in Table 2 relate to what will make linguistic processing easiest for the speaker, those in Table 3 concern what will make it easiest for the hearer. Of course, these need not be mutually exclusive, which explains why the former can appear to be a superset of the latter. But it is important to note that when the speaker selects a linguistic form that helps the hearer decode, there is nothing altruistic about it, for the utterances are all aimed towards the successful manipulation of the speaker's physical world or mental experience. Signi®cantly, however, because the ball is in the hearer's court, the speaker is obliged to anticipate the internal knowledge of formulaic sequences of the
478 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
hearer and to match his/her output to it as closely as possible, something which might, if he/she does not share the same linguistic background, entail greater rather than less processing eort.
Figure 1: The roles of formulaic sequences in bene®ting the speaker Figure 1 shows that, once the two functions of formulaic sequences are viewed in terms of their easing of either the speaker's or the hearer's processing pressures, it is possible to represent them as two intersecting parts of the same strategy. The top part of the diagram represents the role of formulaicity in saving the speaker's processing eort. The bottom part represents its role in supporting the speaker's interactive goals through maximizing the chances of hearer comprehension. The common set of sequences, which simultaneously aid speaker production and hearer comprehension, are the discourse markers. They both anchor the structure of the speaker's output, so that it is easier to sequence the ideas ¯uently, and, simultaneously, signal to the hearer where it will be most appropriate, and inappropriate, to begin a turn and what the overall character of the speaker's message is. It should be emphasized that Figure 1 depicts the speaker's purposes and that it does not, therefore, suggest an intersection between, say,
ALISON WRAY 479
the speaker's and the hearer's own processing agendas; the hearer's agenda is not represented. There is no doubt that the hearer can also derive bene®ts if he/she happens also to have stored a formulaic sequence which the speaker has selected in order to help with production, but this is incidental to the speaker's selection process. Similarly, a speaker can no doubt exercise economics of production as a result of using a hearer-oriented sequence. These two secondary eects together predict that the most ecient and successful communication will occur where the speaker and hearer are very familiar with each others' speech patterns, or indeed share the same microvariety. There are a great many implications to this account, and these are explored in more detail in Wray (1998, 1999, 2000, in press) and Wray and Perkins (2000). In our current context the most important observation is that all formulaic sequences can be characterized in terms of their function in bypassing processing, whether it be the speaker's or the hearer's, or both.
EVALUATING THE ASSUMPTIONS The previous section has oered an account of the functions of formulaic sequences, and this reveals a fundamental diculty for the three teaching approaches reviewed earlier. It was observed at the beginning of the paper that formulaic sequences have been targeted in second language teaching because they seem to hold the key to native-like idiomaticity. By drawing on the con®gurations of words that native speakers really use, the learner can avoid some of the excesses of over-generation described by Pawley and Syder (1983). Willis (1990) goes so far as to provide, as part of the teaching support materials, recordings of native speakers doing the same tasks that are set for the learners. This is to be favoured over the inert `correct' dialogues traditionally presented in textbooks. Potentially, such data oers the learner a chance to hear the turns of phrase used by native speakers to achieve a range of conversation management and interactional functions, such as expressing disagreement, changing the topic, holding the turn while planning the next statement, and so on. It also provides valuable examples of common collocations. Meanwhile, in the second section we saw that Willis (1990), Nattinger and DeCarrico (1992), and Lewis (1993) all embrace a common fundamental assumption regarding the level of attention which can, or should, be given to the internal composition of formulaic sequences. Despite their dierences in emphasis, they all believe that formulaic sequences can be, and indeed should be, viewed as combinations of smaller, recombinable, elements. Furthermore, Lewis and Nattinger and DeCarrico go as far as to suggest that since these elements have been identi®ed within native-like linguistic material, the knowledge of them will be extensive enough partly or fully to replace explicit grammar teaching. This belief is implicitly founded on the evidence from young children acquiring L1 and L2 naturally, even though many of the
480 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
variables in their learning situation are very dierent, and there is only limited evidence that adult learners use formulaic sequences as a resource for extracting grammatical generalizations. These two aimsÐto promote native-like linguistic behaviour and to focus attention on the internal composition of the formulaic sequences which most emulate itÐare contradictory. The preceding section showed that, although almost all formulaic sequences undoubtedly have an internal composition, and in many cases one that is entirely regular to the grammar rules of the language, the whole point of selecting a prefabricated string is to bypass analysis. In selecting prefabricated strings the speaker engages in a trade-o, losing the ¯exibility of novel expression in favour of easier encoding for him/ herself, or easier decoding for the hearer. In the teaching syllabuses, the intended outcome of presenting formulaic sequences to the learner is to make him/her behave in a linguistically more native-like way, but the process by which this is encouraged to occur is the breakdown of the sequences into their constituent components, the very thing that native speakers appear not to do, even though they are capable of it. It is, in short, not native-like voluntarily to activate awarenesss of the internal structure of a formulaic sequence, whether that be through learning a dozen phrases containing the word way, as Willis (1990) promotes, or extracting grammar rules, as Nattinger and DeCarrico (1992) propose in their quest to emulate in adult language teaching the processes of ®rst language acquisition (p. xv). In order to ®nd a solution to this diculty, we need to establish the signi®cance of analyticity in the pursuance of idiomaticity in L2 learning. Can and should it be avoided? If not, is there a way of using it to advantage? We shall consider three issues: ®rst, the extent to which it is possible to emulate the child's changing balance of analytic and holistic processing; second, the consequences of embracing analyticity; and third, the consequences of avoiding analyticity.
The role of analyticity in child and adult language acquisition We saw above that children, both in L1 and L2 acquisition, do seem able to use word strings that they understand only at a holistic level, as input for a process of segmentation and analysis. Yet there is little evidence for such a process in naturalistic L2 acquisition by adults, and only limited indication of analysis in classroom learners. What is it about the child's learning experience that is dierent from the adult's? Wray and Perkins (2000), extending an idea proposed by Locke (1997) and drawing on evidence from studies of the linguistic competence of younger and older children (e.g. Gibbs 1991, Nippold and Martin 1989), propose that `[language development] until roughly the age of 8 . . . is marked by a preference for analytic over formulaic language processing' and that from then until adulthood `the organization of the language system becomes progressively more formulaic' (p. 21). They associate this change in processing
ALISON WRAY 481
strategy with aspects of the child's mental and physical development. This places the explanation squarely in the domain of age rather than stage, and would seem to preclude the possibility that learning in the more mature individual will be subject to the same determining factors. If so, it would be unwise to assume that just because the child can apply analytical processes to derive grammatical and lexical information from formulaic sequences, adults can too. However, one aspect of the child's experience could be roughly emulated in the adult learner. Wray and Perkins suggest that the period of analysis coincides with a time during which the child operates within a `sociointeractional bubble . . . both protected from, and largely impervious to, any need to interact with anyone other than its carers' (p. 22). The eect of the `bubble' is that the child experiences relatively little pressure to engage with the complex demands of interaction in the full range of life situations: you may take a young child to the zoo, the theatre or the Queen's garden party, but the moment-by-moment business of the child will be almost entirely unaected by that wider context, for its own world relates to being fed and kept comfortable by its immediate carers. Wray and Perkins' (2000) proposal is that by being protected from the intellectual and emotional stress of interacting in the world beyond the bubble, the child is able to focus its processing attention on the structure of the language, so that it develops the ability for novel expression. By the time this knowledge has been acquired, a large number of strings in constant use will have become stored as formulaic sequences for ease of access, making language production and comprehension more ¯uent and reliable. This in turn frees up the processing for other priorities, including the dicult and often painful learning of how to operate outside the bubble. As the child develops more of its own life, it ®nds itself in more situations in which it is necessary to say the right thing, ensure acceptance by the group, and manipulate the actions and perceptions of people at dierent social distance and in dierent positions of power. This requires tuning into the speech patterns of others and dealing with language holistically rather than, as previously, analytically. As already noted, the child's `socio-interactional bubble' is a product of its continuing physical dependency and emotional and intellectual immaturity, and so is indeed a product of age rather than stage of language acquisition. However, if adult learners were to ®nd themselves in a similar `bubble'Ða ®nite set of interactional situations over which they had a high level of controlÐthen they, too, might be able to focus on an analysis of the language, before, having mastered its details, adopting a more formulaic approach at the more advanced stages of learning. Arguably, the classroom oers such a bubble. Once learners have established the basic skills required for understanding the language of classroom management, know how to ask for repetition or clari®cation, can gain permission to leave the room and apologize for arriving late, and are able to engage in the generic interaction
482 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
with their peers that is required during activities, they reach a stage of day to day competence not dissimilar to that of the young child with its carers. It is not that there is no more interactional language to learn, but that there is no more that is urgently needed within the standard classroom environment. With very little need to focus attention on the language of the classroom itself, perhaps analytic learning can more easily take place. If this is so, then the best analytic learning will occur in a context in which there are no interactional challenges or surprises. Yet Willis, Nattinger and DeCarrico, and Lewis all introduce multi-word strings because they believe that the classroom `bubble' is too restrictive, and that the learner needs the opportunity to encounter and engage with a wider range of interactional situations. Wray and Perkins' (2000) model suggests that as soon as new interactional situations are foregrounded in teaching, the learner has to make a choice between focusing on the interactional outcomes, which may be at the expense of learning, or on the learning, at the expense of engaging with the interaction. This means that, of all the linguistic material that could be presented for the purposes of analysis, the least suitable are the formulaic sequences used in real interaction. There are, however, limits to the parallel that one can draw between the child's bubble and that of the classroom learner. Even if adult learners can operate within a socio-interactional bubble in the classroom, they do not do so outside. Even if they have never even needed to interact in the real world using the L2, they are aware, from their L1 experience, that there exist subtle dierences between how you request an action from one person rather than another, and they will be sensitive to not knowing, say, how to express condolence, complain in a shop or chastise an acquaintance, without being misconstrued. This contrasts with the child learner, who is largely oblivious of any other world than its own immediate one. The awareness of the adult learner that there is much more to learn, may lie behind the problem which seems to aict adult, but not child, learners: the generation of forms that are grammatical but not idiomatic.
The consequences of embracing analyticity The process of analysis encouraged by the three approaches reviewed here entails generalization. For generalization to be useful and valid, there has to be an underlying system that is predictable. Word-strings presented in a syllabus must, if they are to enable the learner to infer lexical patterns or grammatical rules, be semantically and grammatically regular. It follows that some formulaic sequences, namely those that are non-canonical, metaphorical or archaic, must be excluded. Unfortunately, it is these, the `idioms', that are the epitome of formulaicity for many commentators, and the very focus of some teaching approaches (e.g. Irujo 1986, Milton and Evans 1998). Unless the irregular sequences are excluded, or at least formally ¯agged up, how is the learner to know that it is possible to use large amounts and largely
ALISON WRAY 483
speaking as input for analysis and rule-building, but that at large and by and large will not succumb to that treatment and should not be analysed? Excluding the idioms is one thing. The problem of irregularity does not, however, end there, for the wider phenomenon of idiomaticityÐthe preferred way of saying somethingÐis also no respecter of regularity. Idiomaticity is a natural consequence of storing sequences whole, for when the speaker is confronted with the question, What is the easiest way to express this idea?, the answer will be the same on each occasion. As a result, alternative ways of expressing it will rarely be chosen. Once a string is stored and retrieved whole, however, with no attention paid to its internal structure unless there is some reason to do so, that string may not keep pace with changes in the grammar and lexicon. Moon (1998b) provides several examples of words which never occur outside of a single formulaic sequence, including run amok; by dint of SOMETHING; in high dudgeon; at loggerheads (pp. 78±9), and sequences like director general and rather thee than me retain linguistic features which have passed out of the productive language. With such forms, internal inspection is of little value. Rather than seeing such `fossilization' as unusual, it is important to recognize that it is central to the nature of holistic processing. Thus, the more naturalistic the input given to the learner, the more this is likely to cause diculties, for native speakers appear to use the literal and non-literal and the grammatically canonical and non-canonical with a smooth integration which suggests that they either have no formal recognition of the dierences, or at least do not need to activate it.
The consequences of avoiding analyticity If analyticity is a problem when the input is the formulaic sequence, whose internal composition may not be reliable, then might formulaic sequences be used without encouraging attention to the content? Schmidt (1983) presents the case of Wes, a Japanese immigrant, in the USA. He picked up and created formulaic sequences that enabled him to interact fully, yet the knowledge of these did not seem to improve his grasp of the grammar. This suggests that it may be possible, at least for some learners, to master the form and functional power of prefabricated strings without engaging with what they contain. However, as with Wes this could lead to only limited linguistic success, not least because a signi®cant proportion of the formulaic sequences in the target language are frames with one or more slots for an open class item or for morphological variation. These entail an analytic engagement, during both learning and use, without which we would have no more than a list of unalterable phrases heard, memorized and reproduced. We must, in short, look for a way of accommodating both analyticity and formulaicity.
484 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
Accommodating analyticity and formulaicity David Willis (personal communication) believes that, in classroom learners, some measure of analysis is inevitable and, that being so, it should be encouraged, because the gains in overall command of the language far outstrip the losses in terms of native-like idiomaticity. He acknowledges that this has unintended consequences, but he does not consider these serious: frames like `idea/possibility/chance/danger + OF +-ing' . . . carry with them the seeds of language development . . . [Teaching them] leaves open the possibility of, perhaps even encourages, a sequence like `the wish of +-ing' by analogy with idea/hope/intention etc. This doesn't worry me at all. It's a `mistake' along the lines of `She suggested me to . . . ` or `Can you explain me the problem?'Ða very useful and productive overgeneralisation. It leaves us with the mystery of how learners gradually eliminate these overgeneralisations, but as a pedagogic principle we have no alternative but to encourage creativity (personal communication).
In other words, Willis is prepared to accept the contradiction between his goal, which is to bring learners to a mastery of native-like expression, and his means of achieving this, which entails non-native-like processing behaviour and a tendency to overgeneralize. If we accept this compromise, then the success of using formulaic sequences in teaching depends on the sensitivity of the syllabus designer and the teacher to the complex nature of language in use and, in particular, to the potential for the very idiomaticity of an expression to make it less open to generalization than it may seem at ®rst glance. The more natural the data that are being presented to the learner, the more they need to be subjected to control and guidance in delivery.
CONCLUSION This paper has identi®ed an incongruity within three approaches to teaching formulaic sequences: in order to encourage the development of native-like idiomaticity, a fundamentally analytic approach is promoted, even though the very nature of formulaic sequences seems to be that they are not normally analysed. Willis' solution is to abandon the assumption that the learner is acquiring the same processes as the native speaker, and he legitimizes the classroom learners' inherent desire and ability to analyse, by making it the basis for a principled organization of his material. Rather than denying the arti®ciality of the classroom situation, he embraces and capitalizes upon it, even though it cannot deliver a wholly native-like outcome. The skill of the syllabus writer and teacher lies in adequately juxtaposing the learners' ability to accumulate linguistic repertoire through the observation of language in use, with their predilection to apply conscious analysis. It is observation that will best support the developing sense of what `sounds right' in a given
ALISON WRAY 485
context. But it is analysis that will make up the shortfall between what the classroom context can provide and the creative linguistic knowledge which the learner needs to develop. It would be impossible to present only grammatically and semantically regular sequences as a way of preventing the extrapolation of inappropriate rules, for there is no clear dividing line between what is regular and irregular. The eect of formulaicity on strings is to protect them from all but very occasional internal inspection, and subtle distributional restrictions easily arise as a function of idiomaticity. This makes all formulaic sequences potentially unreliable for analysis. As for overgeneralizationÐthe very learner fault which Pawley and Syder (1983) identi®ed as most intractableÐit seems that this may be an inevitable consequence of classroom teaching. If so, then, despite the bold claims, the introduction of formulaic sequences into classroom teaching cannot easily solve the problem of how to make learners sound idiomatic. (Revised version received December 1999)
NOTES 1 The de®nition encompasses some opportunities for discussion that are beyond the scope of this paper. These include the potential for polymorphemic words to be treated as formulaic sequences, and the in-depth comparison of the processes of comprehension and production. 2 Frequency counts need to be sensitive to the idiolect as well as the agreed preferences of a speech community. One possible diculty with corpus studies at present is that they amalgamate the output of many dierent speakers and writers, and thus represent a corporate variety which does not match the language of any one individual.
3 One oft-cited single case-study of a naturalistic adult learner, Hanania and Gradman's (1977), has been considerably over-valued. Their subject did have some memorized formulaic utterances which `resisted segmentation' and were `associated . . . with particular situations' (p. 79), but Hanania and Gradman speci®cally state that the `memorized utterances were not included in the analysis' (p. 78). 4 This raises the interesting possibility that formulaic language might have predated novel grammatical language in human evolution (Wray 1998, 2000).
REFERENCES Aijmer, K. 1996. Conversational Routines in English. London and New York: Longman. Altenberg, B. 1998. `On the phraseology of spoken English: the evidence of recurrent word combinations' in A. Cowie (ed.): Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press. 101±22. Backus, A. 1999. `Evidence for lexical chunks in insertional codeswitching' in B. Brendemoen, E. Lanza, and E. Ryen (eds.): Language Encounters in Time and Space. Oslo: Novus. 93±109.
Bates, E. and B. MacWhinney. 1987. `Competition, variation and language learning' in B. MacWhinney (ed.): Mechanisms of Language Acquisition. Hillsdale, NJ: Erlbaum. 157±93. Bates, E., I. Bretherton, and L. Snyder. 1988. From First Words to Grammar. Cambridge: Cambridge University Press. Becker, J. 1975. The Phrasal Lexicon. Bolt Beranek & Newman Report no.3081, AI Report no. 28. Biskup, D. 1992. `L1 influence on learners'
486 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
renderings of English collocations: a Polish/ German empirical study' in P. J. L. Arnaud and H. BeÂjoint (eds.): Vocabulary and Applied Linguistics. Basingstoke: Macmillan. 85±93. Bolander, M. 1989. `Prefabs, patterns and rules in interaction? Formulaic speech in adult learners' L2 Swedish' in K. Hyltenstam and L. K. Obler (eds.): Bilingualism across the Lifespan. Cambridge: Cambridge University Press. 73±86. Bolinger, D. 1976. `Meaning and memory.' Forum Linguisticum 1: 1±14. Bretherton, I., S. McNew, L. Snyder, and E. Bates. 1983. `Individual differences at 20 months: analytic and holistic strategies in language acquisition.' Journal of Child Language 10: 293±320. Brown, G. 1990. Listening to Spoken English (2nd edn). London and New York: Longman. Butler, C. S. 1997. `Repeated word combinations in spoken and written text: some implications for functional grammar' in C. S. Butler, R. A. Gatward, J. H. Connolly, and R. M. Vismans (eds.): A Fund of Ideas: Recent Developments in Functional Grammar. Amsterdam: IFOTT, University of Amsterdam. 60±77. Butler, C. S. 1998. `Collocational frameworks in Spanish.' International Journal of Corpus Linguistics 3/1: 1±32. Bygate, M. 1988. `Units of oral expression and language learning.' Applied Linguistics 9/1: 59±82. Chafe, W. L. 1968. `Idiomaticity as an anomaly in the Chomskyan paradigm.' Foundations of Language 4: 109±27. Code, C. 1987. Language, Aphasia, and the Right Hemisphere. Chichester: John Wiley. Code, C. 1997. `Can the right hemisphere speak?' Brain and Language 57: 38±59. Coulmas, F. 1979. `On the sociolinguistic relevance of routine formulae.' Journal of Pragmatics 3: 239±66. Coulmas, F. 1994. `Formulaic language' in R. E. Asher (ed.): Encyclopedia of Language and Linguistics. Oxford: Pergamon. 1292±3. Cowie, A. P. 1988. `Stable and creative aspects of vocabulary use' in R. Carter and M. McCarthy (eds.): Vocabulary and Language Teaching. London and New York: Longman. 126±39. Cowie, A. P. 1994a. `Applied Linguistics: Lexicology' in R. E. Asher (ed.): Encyclopedia
of Language and Linguistics. Oxford: Pergamon. 177±80. Cowie, A. P. 1994b. `Phraseology' in R. E. Asher (ed.): Encyclopedia of Language and Linguistics. Oxford: Pergamon. 3168±71. DeCock, S., S. Granger, G. Leech, and T. McEnery. 1998. `An automated approach to the phrasicon of EFL learners' in S. Granger (ed.): Learner English on Computer. London and New York: Addison Wesley Longman. 67±79. Eeg-Olofsson, M. and B. Altenberg. 1994. `Discontinuous recurrent work combinations in the London-Lund corpus' in U. Fries, G. Tottie, and P. Schneider (eds.): Creating and Using English Language Corpora. Amsterdam: Rodopi. 63±77. Ellis, N. C. 1996. `Sequencing in SLA: phonological memory, chunking and points of order.' Studies in Second Language Acquisition 18: 91±126. Ellis, R. 1984. `Formulaic speech in early classroom second language development' in J. Handscombe, R. A. Orem, and B. P. Taylor (eds.): On TESOL '83. Washington DC: TESOL. 53±65. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford: Oxford University Press. Gibbs, R. W. 1991. `Semantic analyzability in children's understanding of idioms.' Journal of Speech and Hearing Research 34: 613±20. Granger, S. 1998. `Prefabricated patterns in advanced EFL writing: collocations and formulae' in A. Cowie (ed.): Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press. 145±60. Hakuta, K. 1974. `Prefabricated patterns and the emergence of structure in second language acquisition.' Language Learning 24: 287±97. Hanania, E. A. S. and H. L. Gradman. 1977. `Acquisition of English structures: a case study of an adult native speaker of Arabic in an English-speaking environment.' Language Learning 27/1: 75±91. Hatch, E., S. Peck, and J. Wagner-Gough. 1979. `A look at process in child secondlanguage acquisition' in E. Ochs and B. B. Schieffelin (eds.): Developmental Pragmatics. New York: Academic Press. 269±78. Hickey, T. 1993. `Identifying formulas in first language acquisition.' Journal of Child Language 20: 27±41. Hinnenkamp, V. 1980. `The refusal of second
ALISON WRAY 487
language learning in interethnic context' in H. Giles, W. P. Robinson, and P. M. Smith (eds.): Language: Social Psychological Perspectives. Oxford: Pergamon. 179±84. Howarth, P. 1998a. `Phraseology and second language proficiency.' Applied Linguistics 19/ 1: 24±44. Howarth, P. 1998b. `The phraseology of learners' academic writing' in A. P. Cowie (ed.): Phraseology: Theory, Analysis and Applications. Oxford: Clarendon Press. 161±86. Huang, J. and E. M. Hatch. 1978. `A Chinese child's acquisition of English' in E. M. Hatch (ed.): Second Language Acquisition: A Book of Readings. Rowley, MA: Newbury House. 118±31. Hudson, J. 1998. Perspectives on Fixedness: Applied and Theoretical. Lund: Lund University Press. Hughlings Jackson, J. 1874/1958. `On the nature of the duality of the brain' in J. Taylor (ed.): Selected Writings of John Hughlings Jackson, vol 2. London: Staples Press. 129±45. Irujo, S. 1986: `A piece of cake: learning and teaching idioms.' ELT Journal 40/3: 236±42. Krashen, S. and R. Scarcella. 1978. `On routines and patterns in language acquisition and performance.' Language Learning 28/2: 283±300. Kuiper, K. and D. Tan Gek Lin. 1989. `Cultural congruence and conflict in the acquisition of formulae in a second language' in O. Garcia and R. Otheguy (eds.): English Across Cultures: Cultures Across English. Mouton. 281±304. Lattey, E. 1986. `Pragmatic classification of idioms as an aid for the language learner.' IRAL 24/3: 217±33. Lewis, M. 1993. The Lexical Approach. Hove: Teacher Training Publications. Lieven, E. V., J. M. Pine, and H. D. Barnes. 1992. `Individual differences in early vocabulary development: redefining the referential±expressive distinction.' Journal of Child Language 19: 287±310. Locke, J. L. 1993. The Child's Path to Spoken Language. Cambridge, MA: Harvard University Press. Locke, J. L. 1997. `A theory of neurolinguistic development.' Brain and Language 58: 265± 326. McCrone, J. 1999. `States of mind.' New Scientist 2178, 20 March: 30±3.
Milton, J. and V. Evans. 1998. A Good Turn of Phrase: Advanced Idiom Practice. Swansea: Express. Moon, R. 1992. `Textual aspects of fixed expressions in learners' dictionaries' in P. J. L. Arnaud and H. BeÂjoint (eds.): Vocabulary and Applied Linguistics. Basingstoke: Macmillan. 13±27. Moon, R. 1998a. `Frequencies and forms of phrasal lexemes in English' in A. P. Cowie (ed.): Phraseology: Theory, Analysis and Applications. Oxford: Clarendon Press. 79±100. Moon, R. 1998b. Fixed Expressions and Idioms in English. Clarendon Press, Oxford. Myles, F., J. Hooper, and R. Mitchell. 1998. `Rote or rule? Exploring the role of formulaic language in classroom foreign language learning' Language Learning 48/3: 323±63. Myles, F., R. Mitchell, and J. Hooper. 1999. `Interrogative chunks in French L2: a basis for creative construction?' Studies in Second Language Acquisition 21/1: 49±80. Nattinger, J. R. and J. S. DeCarrico. 1992. Lexical Phrases and Language Teaching. Oxford: Oxford University Press. Nelson, K. 1981. `Individual differences in language development: implications for development and language.' Developmental Psychology 17/2: 170±87. Nippold, M. A. and S. T. Martin. 1989. `Idiom interpretation in isolation versus context: a developmental study with adolescents.' Journal of Speech and Hearing Research 32: 59±66. Pawley, A. and F. H. Syder. 1983. `Two puzzles for linguistic theory: nativelike selection and nativelike fluency' in J. C. Richards and R. W. Schmidt (eds.): Language and Communication, New York: Longman. 191± 226. Perkins, M. R. 1994. `Repetitiveness in language disorders: a new analytical procedure.' Clinical Linguistics and Phonetics 8/4: 321±36. Perkins, M. R. 1999. `Productivity and formulaicity in language development' in M. Garman, C. Letts, B. Richards, C. Schelletter, and S. Edwards (eds.): Issues in Normal and Disordered Child Language: From Phonology to Narrative (Special Issue of The New Bulmershe Papers). Reading: University of Reading. 51±67. Peters, A. M. 1977. `Language learning strategies: does the whole equal the sum of the parts?' Language 53/3: 560±73.
488 FORMULAIC SEQUENCES IN SECOND LANGUAGE TEACHING
Peters, A. M. 1983. Units of Language Acquisition. Cambridge: Cambridge University Press. Plunkett, K. 1993. `Lexical segmentation and vocabulary growth in early language acquisition.' Journal of Child Language 20: 43±60. Raichle, M. E. 1998. `The neural correlates of consciousness: an analysis of cognitive skill learning.' Philosophical Transactions of the Royal Society of London, Series B, 353: 1889± 901. Raupach, M. 1984. `Formulae in second language speech production' in H. W. Dechert, D. MoÈhle, and M. Raupach (eds.): Second Language Production. TuÈbingen: Gunter Narr. 114±37. Rehbein, J. 1987. `Multiple formulae: Aspects of Turkish migrant workers' German in intercultural communication' in K. Knapp, W. Enninger, and A. Knapp-Potthoff (eds.): Analysing Intercultural Communication. Berlin: Mouton. 215±48. Reiss, N. 1989. `Speech act taxonomy, chimpanzee communication, and the evolutionary basis of language' in J. Wind, E. G. Pulleybank, E. DeGrolier, and B. H. Bichakjian (eds.): Studies in Language Origins vol. 1. Amsterdam: John Benjamins. 283± 304. Richards, J. C. and T. S. Rodgers. 1986. Approaches and Methods in Language Teaching. Cambridge: Cambridge University Press. Scarcella, R. 1979. ` ``Watch up!'': a study of verbal routines in adults second language performance.' Working Papers in Bilingualism 19: 79±88. Schmidt, R. W. 1983. `Interaction, acculturation, and the acquisition of communicative competence: a case study of an adult' in N. Wolfson and E. Judd (eds.): Sociolinguistics and Language Acquisition. Rowley, MA: Newbury House. 137±74. Sinclair, J. 1991. Corpus, Concordance, Collocation. Oxford: Oxford University Press. Sinclair, J. and A. Renouf. 1988. `A lexical syllabus for language learning' in R. Carter and M. McCarthy (eds.): Vocabulary and Language Teaching. London and New York: Longman. 40±160. Stubbs, M. 1995. `Collocations and cultural connotations of common words.' Linguistics and Education 7/4: 379±90. Stubbs, M. 1997. `Eine Sprache idiomatisch sprechen: Computer, Korpora, kommunika-
tive Kompetenz und Kultur' in K. J. Mattheier (ed.): Norm und Variation. Frankfurt am Main: Peter Lang. 151±67. Van Lancker, D. 1987. `Nonpropositional speech: neurolinguistic studies' in A. W. Ellis (ed.): Progress in the Psychology of Language, vol. 3, Lawrence Erlbaum, Hillsdale, NJ. 49±118. Van Lancker, D. R. and D. Kempler. 1987. `Comprehension of familiar phrases by leftbut not by right-hemisphere damaged patients.' Brain and Language 32: 265±77. Weinert, R. 1995. `The role of formulaic language in second language acquisition: a review.' Applied Linguistics 16/2: 180±205. Weinreich, U. 1969. `Problems in the analysis of idioms' in J. Puhvel (ed.): Substance and Structure of Language. Berkley and Los Angeles: University of California Press. 23±81. Williams, M. 1988. `Language taught for meetings and language used in meetings: is there anything in common?' Applied Linguistics 9/1: 44±58. Willis, D. 1990. The Lexical Syllabus. London: Harper Collins. Willis, D. 1999. `Syllabus design and the pedagogic corpus' in J. Aitchison, H. Funk, R. Galisson, G. List, M. A. Mochet, C. O'Neil, C. Owen, W. Ulrich, G. Vigner, and D. Willis Vocabulary Learning in a Foreign Language. Fontenay/St-Cloud: ENS Editions (Ecole Normale Superieure/ British Council/ Goethe-Institut). 115±48. Wong Fillmore, L. 1976. `The second time around: cognitive and social strategies in second language acquisition' Unpublished Ph.D. thesis, Stanford University. Wray, A. 1990. `The dual systems (``focusing'') hypothesis: a right hemisphere account for left hemisphere language.' Speculations in Science and Technology 13/1: 3±12. Wray, A. 1992. The Focusing Hypothesis: The Theory of left Hemisphere Lateralised Language Re-examined. Amsterdam: John Benjamins. Wray, A. 1998. `Protolanguage as a holistic system for social interaction.' Language and Communication 18: 47±67. Wray, A. 1999. `Formulaic language in learners and native speakers.' Language Teaching 32 (4): 213±31. Wray, A. 2000. `Holistic utterances in protolanguage: the link from primates to humans' in C. Knight, M. Studdert-Kennedy, and J. Hurford (eds.): The Evolutionary Emergence
ALISON WRAY 489
of Language. Cambridge University Press. 285±302. Wray, A. in press. Formulaic Language and the Lexicon. New York: Cambridge University Press. Wray, A. and M. R. Perkins. 2000. `The functions of formulaic language: an integrated model.' Language and Communication 20/1: 1±28.
Yorio, C. A. 1980. `Conventionalized language forms and the development of communicative competence.' TESOL Quarterly 14/4: 433±42. Yorio, C. A. 1989. `Idiomaticity as an indicator of second language proficiency' in K. Hyltenstam and L. K. Obler (eds.): Bilingualism across the Lifespan. Cambridge: Cambridge University Press. 55±72.
Applied Linguistics 21/4: 490±516
# Oxford University Press 2000
Testing L2 Vocabulary Knowledge at a High Level: the Case of the Euralex French Tests PAUL BOGAARDS Leiden University Testing L2 vocabulary knowledge should take into account all aspects of what is implied by lexical knowledge: knowledge of form, meaning, morphology, syntax, collocations, and use in discourse. Dierent types of tests will be necessary to tap all these aspects. Some general aspects of vocabulary testing are brie¯y discussed in this context. In the second part of the paper, a new type of test which aims at testing very high levels of L2 vocabulary knowledge is analysed and investigated. Although the validity of the test as it stands is questionable, it is shown that it is possible to improve on both its reliability and its validity, and to adapt it to the level of vocabulary knowledge which is common with advanced university students.
Testing vocabulary knowledge in a second or foreign language is not as straightforward an aair as is sometimes thought. Depending on what exactly one wants to know about L2 lexical knowledge, one has to select the appropriate materials and adequate procedures to arrive at valid and reliable results. As lexical knowledge comes in very many forms and presents a lot of dierent aspects, this means that there is not one single valid way to measure L2 vocabulary knowledge. Dierent types of tests are needed to address dierent aspects of the lexicon and dierent formats may be more or less adapted to dierent levels of vocabulary knowledge and to dierent types of questions the teacher or the researcher wants to answer. In the ®rst part of this paper I will brie¯y discuss the question of what is implied by `knowing a word' and I will comment on some general problems concerning the design of vocabulary tests. In the second part I will report on a piece of empirical research about one particular form of test, the Euralex French Tests (EFT), which was devised to measure very high levels of lexical knowledge in French as a foreign language. I will give an analysis of the content of these tests and I will present experimental studies on their validity, their reliability and their discriminating power, and I will describe some necessary further developments of the tests. In doing so I will try to demonstrate that an EFT-type of test is potentially a valid, reliable, and useful tool for the measurement of speci®c aspects of lexical knowledge in an L2.
PAUL BOGAARDS
491
TESTS OF LEXICAL KNOWLEDGE IN L2 In order to know what aspects of lexical knowledge can or should be tested, it is important to have a clear view of the nature of what can be known about lexical units. Several attempts have been made at establishing an exhaustive list of aspects of lexical knowledge. Cronbach (1942) distinguishes ®ve aspects: generalization (knowing the de®nition), application (knowledge about use), breadth of meaning (knowing dierent senses of a word), precision of meaning (knowing how to use the word in dierent situations) and availability (knowing how to use the word productively). Richards (1976) gives a list of seven criteria: 1 2 3 4 5 6 7
knowing the degree of probability of encountering that word in speech or print; knowing the limitations imposed on the use of the word according to variations of function and situation; knowing the syntactic behaviour associated with that word; knowledge of the underlying form of a word and the derivations that can be made from it; knowledge of the network of associations between that word and other words in the language; knowing the semantic value of a word; knowing many of the dierent meanings associated with the word.
More recently, Nation (1990: 31) has proposed four dimensions of lexical knowledge: form (oral or written), position (grammar and collocations), function (frequency and appropriateness) and meaning (conceptual and associative). All these forms of knowledge can be productive or receptive. These three approaches to the question what it means to `know a word' are more or less complementary to each other. For instance, Nation adds aspects concerning the knowledge of the written or spoken form of the word, but he does not mention problems of morphology, which are taken into account only by Richards. Whereas Cronbach does not speak about associations or collocations, Nation seems to ignore the possible polysemy of words. To dierent degrees, in these approaches the word is taken as the unit one has knowledge about. It is not speci®ed, however, what is meant by a word. As is well-known, the concept of `word' has never been very clear in linguistic theory, although many dierent de®nitions have been formulated. Generally, such dierent elements as ®nch, which has a single meaning, and ®ne, which has many and belongs to several parts of speech, to, which has above all a grammatical function, and bye, which has rather a pragmatic function, free, which is a simple form, and freedom or freelancer, which are dierent types of complex forms, are all called `words'. On the other hand, elements like give up, heat wave or out with, which may be considered to have the same kind of semantic and grammatical behaviour as the elements called `words' are mostly excluded from that category.
492 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
Following Cruse (1986), I prefer to use the concept of lexical unit, which is assumed to be the union of one stable meaning and a well-de®ned form. According to this de®nition, there are, on the one hand, dierent lexical units sharing the same form: party (political party, my neighbour's party, rescue party, etc.). On the other hand, elements like give up, heat wave and out with as well as spill the beans or Fortune favours fools are considered as lexical units in their own right. As I have tried to make clear elsewhere (Bogaards 1994, 1996, forthcoming), L2 learners may be assumed to learn lexical units, not only `words'. As to the aspects that may be learned about lexical units in an L2, at least the following can be listed: 1 2
3
4
5
formÐLearners have to get acquainted with the written and/or the spoken form of the unit. Knowing that a given form does indeed belong to a given language seems to be a ®rst stage of knowledge. meaningÐKnowledge of the semantic side of a lexical unit may come in dierent shapes. One can have a vague notion, e.g. that haematin has something to do with the blood or that a beech is some kind of tree. It will be clear that knowing the dierences between arrogant, presumptuous, and superior is of another order. In other words, knowing something about the meaning of a lexical unit does not necessarily mean that one knows its meaning nor does it imply that that element has been fully integrated into the semantic network it belongs to or that one has understood all its connotations. Moreover, knowing (something of) one meaning that is associated with some form does not imply knowledge of other meanings that that same form may have, as in the case of party. morphologyÐLexical units have their own conditions on derivation and compounding. Gracefully and graciously have relationships to two dierent lexical units which share the form grace. Even when the morphological mechanisms of the L2 are well understood, many of the actual relationships have to be learned one by one (see also Bogaards 1994: 53±7). Especially for productive use it is dicult for the L2 learner to know whether a given form is possible and in what sense it may be used. syntaxÐA learner who knows the rules of the syntax may make many mistakes by not applying the right rules to the right lexical units. This applies especially to verbs and, to a lesser degree, to adjectives. Learners have to ®nd out how many and what types of arguments are obligatory or possible with a verb in a given sense, or which prepositions have to be used with a verb or an adjective in some speci®c sense, i.e. with a particular lexical unit. collocatesÐWhereas some lexical units, like very or red, seem to be usable in combination with a great number of other elements of a given category, others have a very restricted realm of use. Most of the collocations that seem so natural to native speakers make for great trouble for L2 learners.
PAUL BOGAARDS
6
493
discourseÐIn what type of discourse and to what eect can lexical units like furthermore, moreover and what is more be used? Which lexical units are to be avoided when speaking to someone belonging to another ethnic group or when writing a letter of application? Knowledge of style, register and appropriateness of particular senses of a same word is notoriously dicult to acquire for L2 learners, and it takes a long time before they have a feeling about the relative frequency of lexical units in dierent kinds of contexts.
This list may make it clear that there is much more to testing vocabulary knowledge than just ascertaining that learners know some words. Lexical knowledge is more a matter of degree than of `yes' or `no' and implies far more than just knowing one meaning for each form (Wesche and Paribakht 1996: 14). In addition, all these types of knowledge may be readily accessible and available at any moment for productive use or else be restricted to certain contexts or only recognized when used by others. In the last twenty years, several studies have been devoted to the question of how vocabulary knowledge can or should be measured. In most cases a new test format is discussed and compared to more traditional ones. Such is the case with the Vocabulary Levels Test (Nation 1983), the yes/no-test (Meara and Buxton 1987), the checklist (Read 1988), the C-test (Singleton and Little 1991, Chapelle 1994), the Lexical Frequency Pro®le (Laufer and Nation 1995) or the Vocabulary-size Test of Controlled Productive Ability (Laufer and Nation 1999). Although more general remarks on vocabulary testing are not absent from these studies, more principled discussions about the ways and techniques that can be used to gain insight in the lexical knowledge of L2 learners are harder to ®nd. As an introduction to his Word Associates Test, Read (1993: 355) proposes `four somewhat overlapping dimensions that can be seen to underlie the design of vocabulary test formats'. According to Read, when deciding on a test form, one has to take into account the following aspects: 1 2 3 4
simple or more complex test formats, the yes/no-test being the simplest form; veri®able responses or self-report (i.e. testees have to say whether they know the word or not); size or quality of knowledge, or what is called breadth as opposed to depth of lexical knowledge; testing in context or in isolation.
As can be seen, the ®rst two of Read's dimensions have to do with the type of reaction that is required from the test taker, whereas the third one is ®rst of all concerned with the type of knowledge that is measured and the fourth one is about test conditions. In order to describe the existing test forms which are used for measuring breadth of vocabulary knowledge and as an introduction to their Vocabulary
494 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
Knowledge Scale, Wesche and Paribakht (1996: 15) present the following list of characteristics: 1 2 3 4
how target vocabulary items are selected; the nature of the task presented to the testee; the test response format used; criteria for judging open-ended responses.
In this case the second and the third characteristics seem to overlap with the ®rst two dimensions of Read, whereas the ®rst characteristic adds an important element of test development and the last one takes into account an aspect of test evaluation. Joe, Nation, and Newton (1998 unpublished) describe eight dierent test formats by combining three factors aecting test item diculty. These factors are: 1 2 3
the receptive/productive (or passive/active) distinction; the recognition/recall distinction (roughly the distinction between multiple choice items and ®ll in items); the imprecise/precise distinction.
In each of these distinctions, the ®rst element is said to be easier than the one after the slash. Among these distinctions, the second one is again about type of response, whereas the third factor, as it has to do with the `degree of accuracy required in the answer', seems to have some overlap with the last characteristic of Wesche and Paribakht. As to the ®rst distinction, Joe et al. is the only one of these three attempts at describing important aspects of vocabulary tests to take into account the dierence between receptive and productive knowledge. What these three lists have in common, in spite of their quite dierent aims, is that they try to present rather heterogeneous aspects of testing in one framework: aspects of content, format, construction, and/or evaluation of vocabulary tests tend to be amalgamated. Although these aspects are indeed very much interrelated, it is necessary to keep them conceptually apart, and all the more so in view of the very complex nature of lexical knowledge. From the discussions which the authors mentioned so far devote to existing and new types of vocabulary tests, it appears that tests measuring vocabulary size or breadth of lexical knowledge are the most widespread. Wesche and Paribakht (1996: 26) list twelve dierent formats that can be used to evaluate this aspect and they conclude that these tests `are useful in providing broad characterisations or comparisons of vocabulary size, such as for admission to programs requiring knowledge of the given language or placement in language courses'. That is to say that for practical goals, existing measures of vocabulary size seem acceptable. But at the same time, Meara (1996: 38) states that one of the basic problems in vocabulary research is `that there are no reliable tests of vocabulary size'. On the other hand Wesche and Paribakht (1996: 26) rightly stress that for L2 users other types of tests are also necessary
PAUL BOGAARDS
495
because partial knowledge is important and because `advanced learners need depth and speed of access as well as range in their vocabulary knowledge, for ease and precision of comprehension as well as for eective composition and oral expression'. This brings us to an aspect that is only seldom taken explicitly into account when vocabulary testing is discussed, although testers know that it is the ®rst thing one should be aware of (Hughes 1989: 7, 48): (vocabulary) tests are used for rather dierent purposes. On several occasions it has been suggested that tests of lexical knowledge only answer one question, which is presented as the only one worth answering. Meara (1996: 37), for instance, states that `[t]he basic dimension of lexical competence is size' (see also Read 1988: 14). However, as was stated above, Wesche and Paribakht are not only interested in global measures of vocabulary size, but also in partial knowledge and in depth of knowledge. Recently Read (1997: 311±17) has made an attempt at classifying vocabulary tests according to purpose. Next to forms of integrative lexical measures such as the cloze procedure and the C-test, he tries to make a distinction between vocabulary measures used in a research context or in a school setting. Unfortunately, he seems to mix up these two categories when he mentions Meara's yes/no-test and Nation's Vocabulary Levels Test as useful instruments for classroom teachers and school administrators in the section about research. For several reasons, then, it is important to think about the measuring of other aspects of lexical knowledge. Even if one accepts size or breadth as an essential dimension, this does not imply that other dimensions are unimportant. Especially for advanced learners, it is important that they acquire more senses of polysemous words and learn more about possible collocates, special uses, and so on. Teachers and researchers should be able to investigate progress in vocabulary acquisition, even in situations where no or very few new word forms have been learned. Consistent with this view, Read (1988: 17) argued that particularly for achievement testing `we need to concentrate on depth of knowledge: how well are smaller sets of key vocabulary items known?' Later on he developed the Word Associates Test, a format that involves `a simple response task and allow[s] broad coverage of a set of words while, at the same time, probing depth of knowledge of words in some meaningful way' (Read 1993: 358). In this test, items are composed of a target word and eight other words, four of which have some sort of relationship with the target word and four do not. Testees have to select the four words having a relationship. For a word like edit, related words would be revise (paradigmatic relationship), ®lm (syntagmatic relationship) and publishing (analytic relationship, i.e. a relationship where the related word `represents one aspect, or component, of the meaning of the stimulus and is likely to form part of its dictionary de®nition'). After a pre-test, Read has given two tests of 50 items each to more than 200 learners of English. The tests turned out to be very reliable (KR-20 of .92), the intercorrelation between the two tests was very high (.97) and the
496 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
correlations with a test of words taken from the University Word List were satisfactory (about .75). However, it was found that the associates of some of the stimuli had semantic connections among themselves and verbal reports of some test-takers made the author `cautious about interpreting the successful identi®cation of two or even three associates in an item as necessarily evidence that the test-taker knows the stimulus word. It seems that, frequently, the ability to select correct responses is based on a combination of good vocabulary knowledge and a certain resourcefulness in seeking possible associates, as well as the con®dence to make guesses'. It is questionable whether these reservations disqualify this test format. If its only purpose is to measure how well the selected target items are known, then the test may not do a very good job. But one could be interested also in more general qualitative knowledge of the lexicon. In that case it would be interesting to be able to make a dierence between learners who are successful in identifying two or three associates even without knowing the stimulus word, and those who were not struck by any meaningful relationship between the nine words given in each item. Moreover, `resourcefulness in seeking possible associates' and `con®dence to make guesses' may be seen as negative when one wants to know whether selected relationships are recognized by the learner or not. But in a more general way, such strategies seem to be helpful in normal language use and learners who exploit these means may be said to have richer vocabularies than those who do not. Another test that tries in a somewhat similar vein to tap more qualitative aspects of lexical knowledge, the Euralex French Tests (Meara 1992b, henceforth EFT), does not present the kind of diculty pointed out by Read. As there de®nitely is a need for more types of tests which measure depth of vocabulary knowledge and as, up to now, this particular format has gone unnoticed, I would like to present the EFT and examine to what extent they are useful. In the remainder of this paper I would like to give a detailed description of the EFT, to report on research as to their validity, their reliability and their discriminating power, and to show how they can be further developed.
TESTING HIGH LEVELS OF VOCABULARY KNOWLEDGE IN FRENCH EFT: development and content In the EFT, subjects are presented with a fairly large number of items which consist of two words and where the question to be answered is: are these two words related in some way or not? The items have the following format: 1[] 3[]
pied: grue mot: blanc
2[] 4[]
sarcler: bois jouu: beÂbeÂ
PAUL BOGAARDS
497
The instructions given to the testees are as follows: `You have to decide if you can see an obvious connection between the two words'. And two examples are given, one where there is such a connection (between vache `cow' and lait `milk') and one where there is no relationship (between avion `aeroplane' and eÂcrire `to write'). Each test consists of 60 items, 40 of which present words having some kind of relationship and 20 containing words with no relationship whatsoever (dummy items). In order to give the teacher a big choice, the complete set of EFT contains 50 tests numbered from 101 to 150. Testees should be given four tests per session. This test format was developed to assess the vocabulary knowledge of lexicographers and aimed at a very high level. The format was based on earlier EFL tests by Meara (1992a) which had been successfully used with single word items. As there was little time and money available for the construction of these tests (Meara, personal communication), it was not possible to pre-test the items or to study the validity and the reliability of this new form of test. As can be seen in the items given as examples above, testees have to know quite infrequent items like grue (`crane') and jouu (`chubby-cheeked'), and in that perspective the test can be taken to measure vocabulary size. But the real purpose of the test is to examine whether the test-taker knows that there is a relationship between pied and grue and between jouu and beÂbeÂ, whereas there do not exist any particular relationships between the words given in items 1 and 3. In other words, the (very) advanced learner of French L2 has to demonstrate that he knows the expression faire le pied de grue (`to stand about waiting'), that it is particular of a beÂbe to be jouu and that no such expressions or relations exist in the two other items. This seems indeed to be the kind of knowledge that characterizes a high level of lexical knowledge in French. As Meara (1996: 45) states, `vocabulary size is probably the only dimension of any real importance as long as we are dealing with a small lexicon.' Beyond a threshold of some 5,000 or 6,000 words for English it is important to measure other dimensions. Indeed, to get a good score on the EFT, it does not suce to recognize the words as belonging to the vocabulary of French (like in a yes/no-test) or to know the (or a) meaning of a word (like in most multiple choice tests). These tests are about ®xed expressions, collocates and, as we will see below, about particular relationships like synonymy, antonymy, and hyponymy. Even by giving fairly frequent words like rire and barbe, which are used in the expression rire dans sa barbe (`to laugh to oneself '), the test taps knowledge of another type than most other vocabulary tests. This knowledge stems essentially from the domains 2 (`meaning', and especially knowledge of the semantic network) and 5 (`collocates') of the list of aspects of vocabulary knowledge given in section 1 above. In addition, it checks knowledge of lexical units that are composed of more than one content word like soupe au lait (`quicktempered'). According to the Introduction preceding the EFT, the words used in the tests
498 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
have been selected in two ways. First, rare words were selected by working systematically through a large French dictionary and by picking only those words which were unlikely to be known by L2 speakers with a limited, i.e. non-native knowledge of French. All words that were recognized by a small group of non-native speakers were removed. In this way a list of approximately 300 infrequent words was obtained. These were given to a native speaker of French who for each of them supplied associates which were assumed to be easily recognizable for native speakers. This resulted in the ®rst list, which consisted of 300 rare French words with an associate word for each of them. A second list contains approximately 400 words taken from the FrancËais Fondamental, Premier degre (Gougenheim et al. 1956). A native speaker provided associations for each of the items. These associated words had, again, to be obvious to native speakers, but not obvious to L2 speakers who knew only the basic meaning of the words. A total of 700 associated pairs were produced in this way. A third list was created by randomly pairing items of the two source lists. This list of non-associated word pairs (dummy items) was checked for fortuitous associations. Only the word pairs where no relationships were found were included in the third list. Four types of associations were said to be taken into account. The ®rst is of a semantic nature and includes such relationships as synonymy, hyponymy, and antonymy. The second type is based on what are called selection restrictions such as relationships between a verb and its (preferred) subject or object as in cabrer/cheval (`rear up/horse'). The third type aims at ®xed expressions as is the case in frotter/oreilles (frotter les oreilles aÁ qn means `to box somebody's ears'). The last type concerns cultural aspects, an example of which is jeu/amour because of the title of Marivaux's play Le jeu de l'amour et du hasard. It will be clear that this format is potentially very suitable to measure knowledge of a number of aspects of lexical units. Not only semantic relationships can be tested, but it is possible also to check, for instance, whether subjects recognize special relationships between verbs and their subjects or objects, or between adjectives and nouns or adverbs. Moreover, the knowledge of ®xed expressions can be tested, as well as relevant cultural knowledge. Dierent senses of a polysemous word can very well be tested in this procedure. Items testing knowledge of compounds like heat wave are perfectly conceivable and do indeed appear in the EFT (see below). For English, it is not impossible for knowledge of phrasal verbs like give up to be tested by this type of test. Nevertheless, this format cannot test all aspects of lexical knowledge. Knowledge of morphological, syntactical, or discoursal aspects of lexical units (domains 3, 4 and 6 of the list given in the ®rst section above) seems dicult to tap by this procedure. Finally, it is evident that only recognition, not production of lexical knowledge is measured. In the terms of Read (1993) the EFT's constitute simple tests that produce
PAUL BOGAARDS
499
veri®able responses and tap data about the depth of knowledge of lexical items that are presented in isolation rather than in a normal context. Although the criteria proposed by Wesche and Paribakht (1996) were originally meant to describe tests for breadth of vocabulary knowledge, they can be applied to tests like the EFT as well. The selection of the items was done on the basis of a large dictionary, the task of the testee is to indicate whether or not there is a link between the two words of each item, and the test response format is the one just described; only the fourth dimension described by Wesche and Paribakht does not apply because there are no open-ended responses. According to Joe et al. (1998), the test would be classi®ed as receptive, recognition, and rather precise. In order to know more about the content of the tests, I have analysed six of them. Tests EFT 122±27, which, like all other tests, contained 60 items each, were randomly chosen from the set of 50 tests. This analysis revealed that some of the categories of relationships were somewhat broader than how they were announced in the Introduction. First, the category `selection restrictions' indeed contained cases like appaÃter/ poisson (`to lure/®sh') or deÂgouliner/eau (`to drip/water'), where verbs and their typical objects or subjects are involved. But in many cases there were no such relationships. For instance, in obus/guerre (`shell/war') or gerbe/paysan (`sheaf/ farmer'), the relationships have more to do with the conceptual meaning of the words than with any restrictions in a grammatical sense. Often, one of the words of a pair has to be used or can be used when de®ning the other one, as is the case in touchette/guitare (`fret/guitar') or in corbillard/mort (`hearse/dead'). They seem to be better classi®ed as free or schematic associations or as what Read (1993) calls analytic relationships. In the analyses I will treat the associations as a subcategory of the intended category `selection restrictions'. Secondly, the category `®xed expressions' did not only contain words which formed expressions, like in simple/bonjour (simple comme bonjour means `easy as pie'), but there were a considerable number of words forming compound words like bateau/mouche (bateau-mouche `river boat') or identiteÂ/pieÁce (pieÁce d'identite `identity paper'). Likewise, I will treat these as a subcategory of the category `®xed expressions'. It was not always easy to classify the items into one of the categories. A pair like ciel/chevalier (`heaven/knight') seems to refer to the Bible and has been classi®ed as a cultural item. The pair leÂger/air (`light/air') has been classi®ed as a ®xed expression although the words do at the most form some kind of loose collocation. Some items turned out to be really unclassi®able and, probably, have to be considered as mistakes. One such case is insecte/aaire (`insect/ matter' or `insect/business'). On the other hand, one pair, where, according to the key, no relationship was anticipated, turned out to have one: livre/jungle (`book/jungle') as in Rudyard Kipling's The Jungle Book. In the vast majority of the items only the dictionary forms of the words appeared in the test. In about ten cases derived forms were given instead of canonical forms, like assisteÂe instead of assister or d'armes instead of arme. The numbers of items pertaining
500 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
in each test to the ®ve dierent categories, four types of `positive' items and the dummy items, are given in Table 1.
Table 1. Numbers of items of dierent categories in the six EFT tests
Meaning relationship Synonym Hyponym Antonym Selection restrictions Verbs/nouns Free associations Fixed expressions Expressions Compound words Cultural aspects Unclassi®able Dummy items Total
122
123
124
125
126
127
10 2 Ð
2 1 Ð
9 1 Ð
8 1 Ð
11 1 Ð
9 1 Ð
7 5
2 18
6 5
5 10
5 4
2 12
12 4 Ð Ð 20 60
10 3 4 Ð 20 60
10 4 2 3 20 60
9 4 2 1 20 60
12 6 1 Ð 20 60
12 2 2 Ð 20 60
Total 56 49 7 Ð 81 27 54 88 65 23 11 4 120 360
Table 1 shows that the four types of `positive' items are unevenly represented. There does not seem to be any important reason why they should be represented more evenly, though, ideally, one might wish to ®nd the dierent categories in proportions that would correspond to normal language use. The number of cultural items appears to be quite low, however, and there were no instances of antonyms in this sample. What the table does not show is that several items appeared more than once in the six tests. This was due to the fact that the tests were built up by a random selection from the data base. One item even ®gured in all six tests used for this study. As this was an item featuring a hyponymic relationship, this subcategory was in fact represented by only two instances. The total number of dierent items was 341.
Pre-testing the EFT As a ®rst step to study the content validity of the EFT, the six tests were given to 30 subjects having a superior command of French vocabulary. They were ten teachers of French working in Dutch universities and teacher training colleges, and twenty native speakers of French, some of them university teachers of French in France or in the Netherlands, others lexicologists
PAUL BOGAARDS
501
working in a French dictionary company. The idea was that if the EFT is a valid tool for the measurement of a very high level of French vocabulary knowledge, these native speakers, teachers of French and lexicologists should all obtain near perfect scores. Even if there certainly are dierences between native speakers of a same language, a test of L2 vocabulary knowledge should contain items native speakers can agree upon so that there cannot be much of a discussion about the nature of the relationship between any two words of the test. So, all items should produce scores of at least 90 per cent correct answers. The six tests were copied and were presented to these subjects on an individual basis. The results are given in Table 2. As may be seen, neither of the two groups had a perfect score or even the target minimum score of 90 per cent. In fact the native speakers obtained a score of about 83 per cent and the non-native speakers of about 80 per cent only. This is too low for groups of subjects that may be considered as experts. For that reason the content validity of this test does not seem to be satisfactory. In order to get a clearer picture of the actual content of the tests it is useful to compare the results of the two groups. The dierences between the scores of the two groups of subjects as measured by t-tests were not signi®cant on the
Table 2. Results of the native and non-native groups on the six EFT tests
Category
Native speakers
Non-native speakers
N of (N = 20) items Mean SD scores
(N = 10) Mean SD scores
Meaning relationship Synonyms 49 Hyponyms 7 Selection restrictions Verbs/nouns 27 Free assoc. 54 Fixed expressions Expressions 65 Compounds 23 Culture 11 Unclassi®able 4 Dummy items 120 Total 360
37.95 3.25
%
10.6 77 3.0 46
%
41.90 3.30
4.1 2.8
86 47
n.s. n.s.
21.20 42.70
3.9 8.8
79 79
22.50 40.10
2.7 2.9
83 74
n.s. n.s.
59.25 21.70 9.05 2.15 104.35 300.00
4.8 1.3 1.5 2.7 10.6 27.6
91 94 82 68 87 83
50.30 18.90 7.30 1.60 102.20 286.40
6.8 2.7 1.9 1.7 11.0 20.1
77 82 66 40 85 80
.000 .001 .012 n.s. n.s. n.s.
502 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
six tests taken together, although they were signi®cantly dierent for three of the tests (EFT 123 at the .01 level, and EFT 124 and 127 at the .05 level). The greatest dierence between the scores of the native and the non-native groups appears with the category of ®xed expressions and compounds (p < .001). Notable dierences also appear with culturally bound items (p < .01). The dierences for all other types of items were not signi®cant. The non-natives had a slightly better score on items where some meaning relationship is implied. This may be related to the fact that near native speakers may have less clear meaning boundaries for the words of their foreign language than native speakers and may, therefore, accept more easily associative meaning relationships like coincer/enfermement (`to jam/con®nement'), where all Dutch subjects recognized the intended `synonymy', whereas only half of the French subjects accepted it. This interpretation seems to be all the more acceptable because of the slightly lower score of the non-natives for the cases where they had to decide that no relationship existed between two given words. In more than 80 cases (out of a total of 360), the two groups of subjects gave quite dierent answers. In cases like cuisse/Jupiter (`thigh/Jove'), tesson/bouteille (`shard/bottle'), poule/dent (`chicken/tooth') and avis/population (`notice/ population') the native speakers of French almost unanimously recognized some sort of relationship whereas at most half of the non-natives answered correctly. These items could be taken to discriminate between native and nonnative speakers at an extremely high level, as was the aim of the test. On the other hand, cases like navreÂ/chagrin (`sorry/sorrow') and monobloc/solide (`cast in one piece/solid'), where a relationship was anticipated, as well as tomber/ pipe (`to fall/pipe'), which were unrelated, were correctly answered by all the non-natives but by only half or less of the native speakers. A number of items turned out to be very unsatisfactory. That is to say that in one or both groups half or more of the subjects gave incorrect, or at least unpredicted, answers. This was the case with 26 per cent of the items: 22 out of 56 items with meaning relationships (39.3 per cent), 24 out of 81 selection restriction items (29.6 per cent), 21 out of 88 ®xed expressions and compounds (23.9 per cent), 5 out of 11 cultural items (45.5 per cent) and 21 out of 120 dummy items (17.5 per cent). Some items seem to be too dicult, even for native speakers: only 7 out of 20 knew that a blanchaille (`fry') is a kind of poisson (`®sh') or that there exists a ®xed expression with fois (`time') and lieÁvre (`hare') (courir deux lieÁvres aÁ la fois literally means `chase two hares at the same time' and can be translated as `try to keep two pots on the boil'), and only three knew the word moitir (`impregnate with water'), which is indeed marked as old or technical in the Petit Robert. In other cases, low scores were due to ¯aws in the tests. All French subjects as well as most of the Dutch university teachers recognized the relationship between livre (`book') and jungle (`jungle'), which had not been anticipated, and only three French subjects saw a meaning relationship between amuser (`to amuse') and appaÃt (`bait'), which is rather vague indeed. A case like limoger/ouvrier (`to dismiss/worker') leads to much hesitation; the
PAUL BOGAARDS
503
use of this verb is severely restricted: only when high-ranking ocials or equivalent are ®red, is this verb used; simple workers can only be licencieÂs (`®red'). Some items appeared in more than one test. A comparison of the results of these items shows that in most cases (10 out of 16) there was no or only a very slight variability in the answers: most if not all subjects stuck to the answers they gave the ®rst time they came across a given item. In the other cases, however, there was somewhat more instability, which was to be found among the native speakers as well as among the non-natives, but never among the two groups for one single item. For instance, raclure/deÂbris (`scraping/debris') was positively answered by eight non-native teachers the ®rst time, but by ®ve of them only the second; amorcer/deÂbuter (`begin/start') was recognized as featuring some sort of synonymy by 15 native speakers at its ®rst appearance and by 18 the second time. This seems to stress the need for rather long tests, although it is not clear what the optimal length of this type of test would be. In sum, only 154 of the 341 dierent items the six tests contained had 90 per cent scores in both groups. Approximately half of these `good' items turned out to be dummy items, where 100 per cent scores in both groups were not rare. What this ®rst experiment has clearly shown is that tests like the EFT have to be extensively pre-tested before they can be used with real subjects.
Reliability and discriminating power of two new tests The second experiment aimed at studying the reliability and the discriminating power of this type of test. With the 154 items which seemed to measure, in a valid way, a high level of vocabulary knowledge, I composed two new tests of 60 items each along the same lines as were followed for the original EFT. To do so, I needed 80 positive items and 40 dummy items. For the latter category there was no problem, but for the meaning relationship, the selection restriction and the ®xed expression categories, the criteria had to be somewhat lowered in a limited number of cases. This means that for some items not 18 (i.e. 90 per cent), but only17 out of 20 native speakers had given a correct answer. Table 3 gives an overview of the content of these two new tests. These tests (see Appendix 1) were given to 30 Dutch speaking students who were at dierent levels of study of French at Leiden University. The students had had a minimum of six years of French at secondary level and had been studying French at university level for from one to ®ve years. They were all native speakers of Dutch. Although not all of these students correspond to the target group the original EFT were devised for, it seemed to be interesting to investigate to what extent such tests could distinguish in a reliable way between university students at dierent levels of their study of French. In order to get indications about their level, the students were asked to give some
504 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
Table 3. Numbers of items of dierent categories in the two new tests A Meaning relationship Synonyms Hyponyms Selection restrictions Verbs/nouns Free association Fixed expressions Expressions Compounds Dummy items Total
B
Total 14
6 1
7 Ð 26
3 10
4 9 40
16 4 20 60
15 5 20 60
40 120
supplementary information. Four aspects were taken into account: the year they began their studies, the mean of the grades they had obtained for the exams measuring language pro®ciency, the weekly number of hours they devoted to contacts with French and French speaking people, and the length of their stays in a French speaking country. Table 4 shows some of the results of the two tests. Several analyses were applied to the results of these new tests. In the ®rst place, the correlation between the two tests was at a satisfactory level (.84, p < .001). The reliability of the two tests (Kronbach's alpha) turned out to be .61 for the ®rst test and .82 for the second one. Only the latter ®gure can be considered to be satisfactory. As to the mean scores, which should be between 50 per cent and 90 per cent if the test is supposed to discriminate between subjects, the ®rst test contained 33 items which did not satisfy this criterion, the second one 34. Together the two tests contained 42 items which were too easy to make a dierence and 25 which were too dicult to do so. This conclusion was con®rmed by the calculated item-total correlations (Rit-values, i.e. values of the discriminating power of each item as compared to the whole test). These values were rather low, most of them lying beneath .30 and almost none lying above .50, as would have been satisfactory. Even more disappointingly, a fair number of the Rit-values turned out to be negative, indicating that the better students tended to make more errors on these items and/or that weaker students tended to give the correct answers. These data lead to the conclusion that the discriminating power of the two new tests for the subjects involved is not sucient. It comes as no surprise, then, that no signi®cant correlations have been measured between these tests and the level the students were at, as stated above these levels being
PAUL BOGAARDS
505
Table 4. Results of the students (N = 30) on the two new tests Item type
Mean score
Rit value
Rit2 value
Test A 1 assoc. 3 assoc. 5 expr. 7 syn. 9 Ð
.63 .13 .36 .46 .97
.05 .40 ±.27 .22 ±.04
±.01 .37 .07 .17
11 13 15 17 19
Ð v/n. syn. Ð Ð
.80 .17 .93 .83 .77
±.03 ±.02 ±.08 ±.05 ±.32
21 23 25 27 29
Ð expr. expr. Ð assoc.
.97 .60 .57 .90 .43
±.11 .36 .03 ±.24 .13
31 33 35 37 39
assoc. expr. expr. hyp. Ð
.40 .93 .40 .43 .90
.03 .41 .27 .01
41 43 45 47 49
expr. Ð Ð expr. assoc.
.53 .83 .77 .63 .57
.11 ±.16 ±.14 .36 .43
51 53 55 57 59
Ð expr. Ð assoc. v/n.
.90 .37 .87 .70 .07
±.19 .29 ±.13 .12 .39
Item type
Mean score
Rit value
Rit2 value
2 4 6 8 10
comp. Ð Ð Ð expr.
.53 .83 .90 .90 .90
.11 ±.32 ±.21 ±.19 .22
.21
12 14 16 18 20
expr. comp Ð syn. Ð
.87 .77 .87 .37 .97
.40 .33 ±.24 .36 ±.22
.45 .37
22 24 26 28 30
syn. v/n. syn. expr. assoc.
1.00 .83 .50 .30 .07
.00 .36 .46 .55 .36
.00 .45 .54 .59 .31
.51 .07 .60 .44
32 34 36 38 40
Ð Ð assoc. assoc. Ð
.80 .87 .77 .37 .90
±.03 ±.03 .36 .23 ±.19
.25
42 44 46 48 50
expr. expr. expr. comp. expr.
.43 .40 .13 .47 .47
±.13 .49 .27 .41 .11
.10 .71 .09 .42 .33
52 54 56 58 60
syn. expr. comp. Ð assoc.
.77 .67 .53 .90 .13
.05 .40 .19 ±.41 .25
.16 .49 .19
.33 ±.26
.52 .10 .25
.41 .45 .31 .21 .50
.19
.35
.46 .32
.45
506 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
Table 4. (cont.): Item type Test B 1 Ð 3 expr. 5 syn. 7 syn. 9 expr.
Mean score
Rit value
Rit2 value
1.00 .70 .17 .70 .93
.00 .25 .64 .41 .11
.29 .61 .47 .10
11 13 15 17 19
Ð Ð expr. v/n. Ð
.90 .60 .17 .07 .97
±.20 ±.18 .40 .57 ±.28
21 23 25 27 29
assoc. expr. Ð Ð Ð
.57 ±.03 ±.17 ±.28 ±.22
31 33 35 37 39
v/n. v/n. Ð Ð syn.
.17 .93 .97 .93 .93 .17
41 43 45 47 49 51 53 55 57 59
.52 .58 .59 ±.01
.40
.13 .80 .73 .63
.43 .41 ±.19 .16 .48
expr. expr. Ð expr. assoc.
.53 .30 .87 .50 .67
.09 .30 .23 .57 .40
.05 .40
syn. expr. Ð expr. Ð
.03 .40 .87 .23 .87
.56 .45 .01 .43 ±.25
.44 .56
.42
.53
.58 .37
.44
Item type
Mean score
Rit value
Rit2 value
2 4 6 8 10
expr. Ð Ð expr. assoc.
.47 .97 .97 .77 .07
.24 ±.28 ±.11 .39 .57
.32
12 14 16 18 20
syn. comp. v/n. expr. expr.
.07 .27 .60 .23 .10
.49 .41 .43 .01 .50
.39 .41 .47 .02 .49
22 24 26 28 30
syn. assoc. assoc. comp. assoc.
.68 .13 .61 .20 .35
.69 .17 .60 .26 .42
32 34 36 38 40
expr. assoc. comp. comp. comp.
.60 .97 .20 .27 .73 .57 .37 .57 .83 .67
.36 .63 .21 .42 .19
.34 .66 .33 .46 .17
42 44 46 48 50
Ð assoc. Ð Ð expr.
.83 .23 .37 .80 .80
.18 .11 .09 ±.38 .45
52 54 56 58 60
Ð syn. assoc. Ð Ð
.83 .60 .33 .83 .97
.02 .55 .58 ±.23 ±.17
.41 .58
.12
.49 .60 .61
(comp. = compound word; assoc. = free associations; expr. = ®xed expression; hyp. = hyponym; syn. = synonym; v/n = selection restriction; Ð = no relationship)
PAUL BOGAARDS
507
measured as the grades they had got for their language pro®ciency exams, the number of years they had been studying French and the length of their stay in a francophone country. Only the second test, which had a higher reliability as measured by Kronbach's alpha, gave some signi®cant correlations with the number of hours per week students declared to pass in reading or listening to French (resp. .51 and .53, p < .01). As to the negative Rit-values, they almost all appear with the dummy items, i.e. the items where no relationship was anticipated (32 out of 40), whereas negative values are almost absent with the positive items (5 out of 80). In order to ®nd an explanation for the negative Rit-values with 32 of the 40 dummy items, it seems sucient to look at cases like pelle/nez (`shovel/nose', test A, item 58), chaux/hamecËon (`lime/®sh hook', A6) or embrayer/deÂsir (`to put into gear/desire', B29) which all combine a very high mean score (90 per cent or more) with a negative Rit-value. In all these items, at least one word does not belong to the average vocabulary of university students of French. And when one word is unknown, there cannot be a relationship for the student; so, the answer will be `no', which in these cases is correct. As more words will be unknown to the low scorers, these students will give more `no' answers, which considerably lowers the discriminating power of this type of item. It should be added however that a combination of a high mean score and a negative Rit-value may also appear with items which contain two very frequent words like agreÂable/dents (`pleasant/teeth', A4) or corde/facile (`rope/ easy', B4). But here again, those who are aware of fewer relationships between words, the low scorers, will tend to give a correct negative answer. The negative Rit-values of the ®ve positive items are quite low, with the exception of one item, A5, which tests knowledge of the expression avoir d'autres chats aÁ fouetter (`to have other ®sh to fry'). Looking at the satisfactory Rit-values (.40 and up), it becomes clear that in almost all cases less technical words are involved. Consider the following examples, presented in order of diminishing Rit-values: panser/soigner (`to put a dressing/to nurse', B22), atout/carte (`trump/card', B34), agrafeuse/papier (`stapler/paper', B26), sillon/champ (`furrow/®eld', B56), pas/loup (`step/ wolf ', B47), obus/guerre (`shell/war', B21), charrue/boeuf (`plough/ox', A28), houille/mineur (`coal/miner', A31). Unfortunately, only some of these items have at the same time acceptable mean scores (between 50 per cent and 90 per cent). As was said above, many items do not satisfy the normal psychometric criteria. In fact, only 18 items combine a mean score value between 50 per cent and 90 per cent and a Rit-value of over .40. An item like vitre/feneÃtre (`pane/window', A22), which has a mean score of 100 per cent and, as a consequence, a Rit-value of .00, is de®nitely too easy and should be removed from the test. The same holds for demander/graÃce (`ask/mercy') with a mean score of 93 per cent and a Rit-value of ±.03. On the other hand, cabrer/cheval (`to rear up/horse', A13) with a mean score of 13 per cent and a Rit-value of only .27 (or .09 after removal of the dummy items, see hereafter) seems to be
508 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
too dicult for the population and does not discriminate in a satisfactory manner either. It is less clear what the value is of an item like tamiser/lumieÁre (`to ®lter/light', A59) which combines a mean score of 7 per cent with a Ritvalue of .39 (but .50 after removal of the dummy items). Such items seem to be too dicult, but discriminate in a rather satisfactory way between students. As the dummy items play a negative role as far as the discriminatory power of the test is concerned, a new analysis of the results was done with only the 40 positive items per test (those with some kind of relationship between the words of a pair). The new Rit-values of these 80 positive items are given in the column Rit-2 value in Table 4. As can be seen, there remain only three negative values, two of which are very low. This means that most of the positive items of the two tests measure the same kind of knowledge. This is indicated also by Kronbach's alpha: .84 for the ®rst test, .90 for the second one. If the two tests are taken together, Kronbach's alpha is at .93. All these ®gures are satisfactory and suggest that two tests of 40 positive items each are sucient to measure knowledge of lexical relationships of university students of French in a reliable way. It should be kept in mind, however, that such a test, as its items do not represent a real sample of French, is not necessarily a test that covers the whole lexicon of French. For the two reduced tests taken together, there was a low but signi®cant correlation with the mean grades the students had got for their language pro®ciency exams (.46, p < .01). Correlations with year of study or time spent in a francophone environment were not signi®cant. It goes without saying that not taking the dummy items into consideration creates a problem for the use of this type of test. If the subjects know how the results will be analysed, all they have to do is to put a plus in all cases. So, either one has to make sure that the subjects do not know which statistics will be applied to their results, or to think of some more sophisticated way of evaluating the results. I will come back to this issue in the next section.
The third test In order to further analyse the usefulness of the test format inaugurated by the EFT, the best items of the two tests analysed in the preceding section were chosen to make another test; the third test. As to the positive items, no items were chosen with negative Rit-values, or with Rit-values lower than .10. All items had to have mean scores between 50 per cent and 90 per cent, or, in case of Rit-values above .20, between 40 per cent and 90 per cent. The dummy items had to have a mean score between 50 per cent and 90 per cent and Ritvalues of ±.15 or above. This leads to a test of 70 items, 58 positive ones and 12 negative ones (see Appendix 2). This test was given to 29 students who were at dierent stages of their study of French at the same Dutch University. Table 5 presents the results of the third test. As can be seen, 19 of the 58 positive items have satisfactory mean
PAUL BOGAARDS
509
scores (between 50 per cent and 90 per cent) and another 10 have acceptable mean scores (between 40 per cent and 50 per cent). There are no mean scores higher than 90 per cent, which means that none of the items must be considered to be too easy for the subjects involved. Half of the positive items have mean scores which lie below 40 per cent. As to the 12 dummy items, 6 have satisfactory mean scores (between 50 per cent and 90 per cent) and 6 have mean scores of more than 90 per cent, two of them giving a 100 per cent score. The overall mean of the test is .50. This means that the test as a whole is rather dicult for the subjects. Table 5 also shows that 24 out of 58 positive items have Rit-values of at least .40, whereas another 14 are superior to .30. Four positive items have negative Rit-values, three of them being near to zero and only one (item 66) behaving in a rather unexpected way: sourd/pot (`deaf/pot': as deaf as a post) had been one of the better items in the second test. Four of the 12 dummy items have negative Rit-values, two of which (items 33 and 54) seem to be of some importance (±.22 and ±.31 respectively). The overall reliability of the test as measured by Kronbach's alpha is .84, which means that the test measures vocabulary knowledge in a consistent way. As was the case with the two tests which were discussed above in the section entitled `Reliability and discriminating power of the two new tests', the results of the third test were correlated with other data which had been provided by the subjects, such as year of study, grades for translation, grammar, speaking and listening and contacts with French and the French. There were signi®cant correlations with `year of study' and `grammar' (.44 and .46 resp., p < .05); there was a highly signi®cant correlation with speaking skill (.61, p < .01). The correlation with translation French±Dutch was at .37 (p = .09). Correlations with translation Dutch±French and listening or with time spent in a francophone environment were not signi®cant. Although most of the correlations are not high, as a whole, the third test can be taken to be a better predictor of language pro®ciency than the two tests of the previous section. In order to examine the role of the dummy items, some additional statistics were computed. In the ®rst place, results were calculated of the positive items only, leaving out all dummy items. This lowered all correlations with the language pro®ciency data, year of study and contact with French with one to ®ve points. This means that, in spite of the negative Rit-values of a number of dummy items, on the whole this type of item adds to the discriminating power of the test. On the contrary, subtracting points for all positively (i.e. erroneously) answered dummy items had a positive eect on the level of correlations. After having compared several weights for the dummy items, counting each positively answered dummy item as two errors gave the best results with a highly signi®cant correlation of .64 with speaking skill. This suggests a solution for the scoring problem. If test takers know that all positively answered dummy items will be counted as two errors, they will have to choose carefully the ones they consider to have no relationship. As to
510 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
Table 5. Results of the students (N = 29) on the third test
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Item type
Mean score
Rit value
comp. assoc. syn. syn. Ð expr. v/n comp. Ð syn expr. v/n expr. assoc. assoc. Ð Ð Ð expr. assoc. hyp. assoc. expr. expr. expr. comp. assoc. expr. syn. expr. expr. expr. Ð comp. assoc.
.45 .07 .24 .41 .72 .83 .07 .83 1.00 .28 .66 .76 .52 .21 .24 .93 .90 .90 .38 .66 .38 .14 .48 .48 .62 .34 .66 .21 .59 .52 .17 .45 .90 .72 .31
.33 .62 .33 .11 .07 ±.06 .60 .01 .00 .58 .27 .26 .48 .27 .31 .02 .00 .00 .35 .44 .28 .61 .40 .21 .32 .44 .46 .55 .44 .27 .43 .34 ±.22 .33 .61
36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
Item type
Mean score
Rit value
assoc. expr. syn. syn expr. comp. assoc. expr. v/n assoc. syn. comp. assoc. v/n expr. v/n assoc. comp. Ð Ð comp. syn. comp. expr. Ð Ð expr. assoc. expr. Ð expr. expr. syn. Ð assoc.
.17 .48 .31 .55 .48 .24 .35 .10 .34 .31 .52 .76 .72 .24 .48 .07 .28 .59 .93 .86 .76 .72 .52 .10 .97 .52 .28 .55 .38 1.00 .10 .34 .45 .93 .24
.35 .36 .48 .32 .33 .41 .42 .42 .31 .4 .25 .44 .29 ±.08 .48 .17 .34 .16 ±.31 ±.05 .27 .37 ±.05 .42 ±.25 .10 .27 .44 .50 .00 ±.29 .22 .46 .34 .45
(comp. = compound word; assoc. = free association; expr. = ®xed expression; hyp. = hyponym; syn. = synonym; v/n = selection restriction; Ð = no relationship)
PAUL BOGAARDS
511
test construction, it seems advisable to have a proportion of one third of dummy items in a test, as was the case in the original EFT.
DISCUSSION AND CONCLUSION Up to now, no entirely satisfactory tests of vocabulary knowledge have been devised. Read (1988: 17) comes to the conclusion that checklists, multiple choice tests as well as tests where words are to be matched with their de®nitions are all formats which can be `criticized as inadequate indicators of whether the word is really known', and Meara (1996: 38) expresses the same opinion. It should be remembered that both authors are speaking about measures of vocabulary size, the dimension that seems to be the easier one when it comes to measuring lexical knowledge. In the meantime it is recognized that there are other aspects that should be paid attention to and that there is a need for `vocabulary measures of diering sensitivity, so that each word is tested several times in dierent ways' (Joe et al. 1998; see also Wesche and Paribakht 1996). Meara (1996: 49) suggests that `it might be possible to develop an alternative way of characterising a lexicon, one that is largely, though not entirely, independent of size', i.e. in terms of organization. The most concrete attempt at measuring quality rather than size, or depth rather than breadth, is Read (1993). But at the end of his paper, Read declares that he is not entirely certain about the extent to which `this format has achieved the goal of measuring depth of knowledge of particular words' (see also Read 1997: 317). One of the major problems in the construction of vocabulary tests is the absence of any standard. In the ®rst place there is a `lack of agreement among theoreticians, researchers and questionnaire respondents about what it means to `know' a word', as Wesche and Paribakht (1996: 14) put it. And they go on (p. 32) by saying that there is, on the one hand, a `lack of theoretical consensus about the nature and course of development of L2 vocabulary knowledge' and, on the other hand a `lack of alternative instruments that aim to measure the level of knowledge of individual words [which] makes it dicult to estimate concurrent validity'. So, on the theoretical level as well as on the practical level, we are confronted with an empty space as far as vocabulary acquisition is concerned. This situation could easily turn into a vicious circle where everyone is waiting for the others: those who want to de®ne vocabulary knowledge want to be able to measure it, but at the same time test constructors will only be able to develop valid and reliable tools if it is clear what has to be understood by vocabulary knowledge. It is in this context that every new attempt at measuring aspects of lexical knowledge must be welcomed and taken seriously. That is what has been done in this paper. And it will be clear that nothing very de®nitive can be said about the EFT or the ideas they were built on. However, what seems clear is that the EFT tests in their original state are not suciently valid: subjects who are known to have the level of vocabulary knowledge required to get high
512 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
scores on these tests do not obtain (near) perfect scores. However, two types of items, those which measure knowledge of ®xed expressions and those where some kind of cultural knowledge is involved, do discriminate between native speakers and very high pro®ciency non-native speakers like university teachers of French. Two tests developed with the best items of six of the EFT tests did not turn out to be suciently reliable tools to measure the vocabulary knowledge of university students. The overall reliability of only one of the tests was satisfactory. A further analysis of the tests revealed that the Rit-values of the dummy items were negative in the vast majority of the cases, meaning that these items measured something other than what was being measured by the rest of the test. If the dummy items are removed, more homogeneous tests result, which give better overall reliabilities (Kronbach's alpha at .84 and .90 respectively). When the two tests without dummy items are taken together, they give a highly reliable image of vocabulary knowledge (Kronbach's alpha at .93) of university students of French. They have, in addition, a modest but signi®cant correlation with grades obtained for pro®ciency in the language. Concerning the latter point, it is dicult to say what the desirable level of the correlations between a test of qualitative lexical knowledge and marks for grammar or translations would be. One of the problems is that we do not know to what degree general language pro®ciency is determined by vocabulary knowledge. So, for the moment, positive correlations with speaking, listening, reading or writing may be taken to be in favour of the content of a vocabulary test. If one or more of these correlations are (highly) signi®cant, this could shed light on the (type of) vocabulary knowledge measured as well as on the special link with the particular type of pro®ciency. However, the tests that have been discussed in this paper do not seem to be useful as practical tools for measuring levels of vocabulary knowledge in an educational context when the dummy items are taken out or when these items are not taken into account in the analysis of the results. Moreover, ignoring the dummy items means that potentially useful information about the lexical knowledge of the subjects is not considered. All this means that it is important to maintain the format of the tests and to ®nd a solution for the statistical treatment of the dummy items. In the light of this argument, a third test was composed with the best items of the two new tests discussed above. As the reliability of the test as a whole, i.e. including the dummy items, was at a satisfactory level, it can be said that this test measures in a consistent way some aspects of depth of lexical knowledge on a fairly high level. As counting positively answered dummy items as two errors had a positive eect on the correlations with measures for language pro®ciency and contact with the language, this method could be used in educational contexts. This method hints at a proportion of one third of dummy items in a test. This paper has tried to demonstrate that an EFT-type of test is potentially a valid, reliable and useful tool for the measurement of speci®c aspects of lexical
PAUL BOGAARDS
513
knowledge in an L2, especially at rather high levels. It has shown that before using this type of test it is important to ascertain its validity by pre-testing the items with native speakers of the language and/or with other groups of speakers who de®nitely have the level of knowledge that is going to be measured. Next, it has been pointed out that it is necessary to study the reliability of the tests by taking into account psychometric data about the items. With the help of these data it is possible to compose tests which measure the relevant knowledge in a reliable way and which are able to discriminate between groups of students. By analysing the results obtained and by improving step by step its content, this type of test, which was at ®rst meant for a restricted set of non-native speakers of French with a very high pro®ciency level, seems to be fairly adaptable to dierent testing environments. As choices can be made concerning the aspects of vocabulary knowledge that one wants to test, it seems worthwhile to carry on research on this format. Final version received January 2000
ACKNOWLEDGEMENTS I would like to thank Dr Lydius Nienhuis (University of Utrecht) and Dr Vincent van Heuven (Leiden University) for their help with the statistics.
APPENDIX 1ÐTHE TWO NEW TESTS Test A isoloir: eÂlection auÃt: chasseur chat: fouetter barbe: ennui aviron: singe
[ [ [ [ [
] ] ] ] ]
2. 4. 6. 8. 10.
bateau: mouche agreÂable: dents chaux: hamecËon poteleÂ: corde dõÃner: chandelle
[ [ [ [ [
] ] ] ] ]
11. 13. 15. 17. 19.
baÃtisse: beÂbe cabrer: cheval deÂcorer: meÂdaille allumer: petit amuser: graÃce
[ [ [ [ [
] ] ] ] ]
12. 14. 16. 18. 20.
bouillir: coleÁre champ: honneur deÂlabreÂ: langue amorcer: deÂbuter enluminure: manger
[ [ [ [ [
] ] ] ] ]
21. 23. 25. 27. 29.
¯eÂtrir: multiplier vue: nez frotter: oreilles champ: clou gerbe: paysan
[ [ [ [ [
] ] ] ] ]
22. 24. 26. 28. 30.
vitre: feneÃtre former: jeunesse gaÃchis: gaspiller charrue: boeuf haÃle: soleil
[ [ [ [ [
] ] ] ] ]
1. 3. 5. 7. 9.
31. houille: mineur 33. jour: ouvrable 35. coq: gaulois
[] [] []
32. ideÂe: pieÁce 34. ciel: bec 36. leÂzarde: mur
[] [] []
514 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
37. babouin: singe 39. manger: court
[] []
38. bail: appartement 40. mentir: poule
[] []
41. 43. 45. 47. 49.
monde: fou octroi: sacre parent: tabac doigt: oeil embolie: sang
[ [ [ [ [
] ] ] ] ]
42. 44. 46. 48. 50.
nez: moutarde oreilles: casser deÂlier: langue eau: javel peau: neuve
[ [ [ [ [
] ] ] ] ]
51. 53. 55. 57. 59.
plein: corde rire: barbe souÃl: hypocrite pelage: animal tamiser: lumieÁre
[ [ [ [ [
] ] ] ] ]
52. 54. 56. 58. 60.
deÂcorer: honneur simple: bonjour soupe: lait pelle: nez vanne: robinet
[ [ [ [ [
] ] ] ] ]
bifteck: mer coleÁre: piquer crinieÁre: chevelure journal: canard meÂdaille: revers
[ [ [ [ [
] ] ] ] ]
2. 4. 6. 8. 10.
blanc: yeux corde: facile jaunisse: panier matin: petit miteÂ: veÃtement
[ [ [ [ [
] ] ] ] ]
11. 13. 15. 17. 19.
boudin: insister chien: chevalier jaune: rire deÂgouliner: eau nord: casser
[ [ [ [ [
] ] ] ] ]
12. 14. 16. 18. 20.
chanson: rengaine civet: lapin croquer: pomme passer: tabac nord: perdre
[ [ [ [ [
] ] ] ] ]
21. 23. 25. 27. 29.
obus: guerre demander: graÃce abside: bois fondre: poignet embrayer: deÂsir
[ [ [ [ [
] ] ] ] ]
22. 24. 26. 28. 30.
panser: soigner deÂmarrage: voiture agrafeuse: papier fosseÂ: geÂneÂrations eÂplucheur: pommes de terre
[ [ [ [ [
] ] ] ] ]
31. 33. 35. 37. 39.
feÃler: craÃne appaÃter: poisson fer: chien blouse: argent fouiller: recherches
[ [ [ [ [
] ] ] ] ]
32. 34. 36. 38. 40.
fer: croiser atout: carte feu: joie boõÃte: nuit identiteÂ: pieÁce
[ [ [ [ [
] ] ] ] ]
41. 43. 45. 47. 49.
induire: erreur mot: maÃcher mouton: aÃge pas: loup deÂvisser: tourne-vis
[ [ [ [ [
] ] ] ] ]
42. 44. 46. 48. 50.
jambe: amour moulinet: peÃche navreÂ: lac escamotage: mer doux: mot
[ [ [ [ [
] ] ] ] ]
51. 53. 55. 57. 59.
pitre: amusant puce: oreille sauter: drapeau sourd: pot peur: pouce
[ [ [ [ [
] ] ] ] ]
52. 54. 56. 58. 60.
poule: image reÂtreÂcir: diminuer sillon: champ tambour: voile pieÁtre: neige
[ [ [ [ [
] ] ] ] ]
Test B 1. 3. 5. 7. 9.
PAUL BOGAARDS
515
APPENDIX 2ÐTHE THIRD TEST 1. 3. 5. 7. 9.
bateau: mouche barbe: ennui baÃtisse: beÂbe cabrer: cheval allumer: petit
[ [ [ [ [
] ] ] ] ]
2. 4. 6. 8. 10.
auÃt: chasseur gaÃchis: gaspiller bouillir: coleÁre champ: honneur amorcer: deÂbuter
[ [ [ [ [
] ] ] ] ]
11. 13. 15. 17. 19.
vue: nez charrue: boeuf houille: mineur ciel: bec coq: gaulois
[ [ [ [ [
] ] ] ] ]
12. 14. 16. 18. 20.
former: jeunesse gerbe: paysan ideÂe: pieÁce manger: court leÂzarde: mur
[ [ [ [ [
] ] ] ] ]
21. 23. 25. 27. 29.
babouin: singe monde: fou doigt: oeil embolie: sang deÂcorer: honneur
[ [ [ [ [
] ] ] ] ]
22. 24. 26. 28. 30.
bail: appartement oreilles: casser eau: javel peau: neuve blanc: yeux
[ [ [ [ [
] ] ] ] ]
31. 33. 35. 37. 39.
rire: barbe souÃl: hypocrite pelage: animal coleÁre: piquer journal: canard
[ [ [ [ [
] ] ] ] ]
32. 34. 36. 38. 40.
simple: bonjour soupe: lait vanne: robinet crinieÁre: chevelure matin: petit
[ [ [ [ [
] ] ] ] ]
41. 43. 45. 47. 49.
civet: lapin jaune: rire obus: guerre fosseÂ: geÂneÂration feÃler: craÃne
[ [ [ [ [
] ] ] ] ]
42. 44. 46. 48. 50.
agrafeuse: papier croquer: pomme panser: soigner eÂplucheur: pomme de terre fer: croiser
[ [ [ [ [
] ] ] ] ]
51. 53. 55. 57. 59.
appaÃter: poisson feu: joie blouse: argent fouiller: recherches mot: maÃcher
[ [ [ [ [
] ] ] ] ]
52. 54. 56. 58. 60.
atout: carte jambe: amour boõÃte: nuit identiteÂ: pieÁce mouton: aÃge
[ [ [ [ [
] ] ] ] ]
61. 63. 65. 67. 69.
navreÂ: lac deÂvisser: tourne-vis poule: image puce: oreille sauter: drapeau
[ [ [ [ [
] ] ] ] ]
62. 64. 66. 68. 70.
pas: loup doux: mot sourd: pot reÂtreÂcir: diminuer sillon: champ
[ [ [ [ [
] ] ] ] ]
516 TESTING L2 VOCABULARY KNOWLEDGE AT A HIGH LEVEL
REFERENCES Bogaards, P. 1994. Le vocabulaire dans l'apprentissage des langues eÂtrangeÁres. Paris: CreÂdif± Hatier/Didier (Coll. LAL). Bogaards, P. 1996. `Dictionaries for learners of English.' International Journal of Lexicography 9: 277±320. Bogaards, P. (forthcoming). `Lexical units and the learning of foreign language vocabulary.' Studies in Second Language Acquisition. Chapelle, C. A. 1994. `Are C-tests valid measures for L2 vocabulary research?.' Second Language Research 10: 157±87. Cronbach, L. J. 1942. `An analysis of techniques for diagnostic vocabulary testing.' Journal of Educational Research 36: 206±17. Cruse, D. A. 1986. Lexical Semantics. Cambridge: Cambridge University Press. Gougenheim, G., P. Rivenc, R. MicheÂa, and A. Sauvageot. 1956. L'eÂlaboration du francËais fondamental (1er degreÂ). Paris: Didier. Hughes, A. 1989. Testing for Language Teachers. Cambridge: Cambridge University Press. Joe, A., P. Nation, and J. Newton. 1998. `Vocabulary test difficulty and item types' unpublished paper. Laufer, B. and P. Nation. 1995. `Vocabulary size and use: Lexical richness in L2 written production.' Applied Linguistics 16: 307±22. Laufer, B. and P. Nation. 1999. `A vocabularysize test of controlled productive ability'. Language Testing 16: 33±51. Meara, P. 1992a. EFL Vocabulary Tests. Swansea: Centre for Applied Language Studies. Meara, P. 1992b. Euralex French Tests. Swansea: Centre for Applied Language Studies.
Meara, P. 1996. `The dimensions of lexical competence' in G. Brown, K. Malmkjaer, and J. Williams (eds.): Performance and Competence in Second Language Acquisition. Cambridge: Cambridge University Press. 35±53. Meara, P. and B. Buxton. 1987. `An alternative to multiple choice vocabulary tests.' Language Testing 4: 142±54. Nation, P. 1983. `Testing and teaching vocabulary.' Guidelines 5: 12±25. Nation, P. 1990. Teaching and Learning Vocabulary. New York: Newbury House. Read, J. 1988. `Measuring the vocabulary knowledge of second language learners.' RELC Journal 19/2: 13±26. Read, J. 1993. `The development of a new measure of L2 vocabulary knowledge.' Language Testing 10: 355±71. Read, J. 1997. `Vocabulary and testing' in N. Schmitt and M. McCarthy (eds.): Vocabulary. Description, Acquisition and Pedagogy, Cambridge: Cambridge University Press. 303±20. Richards, J. 1976. `The role of vocabulary teaching.' TESOL Quarterly 10: 77±89. Singleton, D. and D. Little. 1991. `The second language lexicon: some evidence from university-level learners of French and German.' Second Language Research 7: 62±81. Wesche, M. and T. S. Paribakht. 1996. `Assessing second language vocabulary knowledge: depth versus breadth.' The Canadian Modern Language Review 53: 13±40.
Applied Linguistics 21/4: 517±552
# Oxford University Press 2000
Investigating Instrument-based Pragmatic Variability: Eects of Enhancing Discourse Completion Tests KRISTINE BILLMYER and MANKA VARGHESE University of Pennsylvania A signi®cant dilemma in sociolinguistic research concerns the methods used to collect data, the validity of dierent types of data and, `. . .their adequacy to approximate the authentic performance of linguistic action' (Kasper and Dahl 1991: 215). Manes and Wolfson (1981) claimed that the most authentic data in sociolinguistic research is spontaneous speech gathered by ethnographic observation. Diculties in relying on this method are well-documented (Blum-Kulka et al. 1989, Aston 1995) and have led to the wide use of an elicitation procedure known as the discourse completion test (DCT). Justi®ed criticisms have been levelled at the DCT, some labelling it an instrument that limits the capturing of authentic communication, and others making it look almost obsolete. However, there are to date no other sociolinguistic data collection instruments that have as many administrative advantages as the DCT, making it, practically speaking, a resource pragmatics testing and teaching will continue to rely on. Furthermore, a better understanding of communication in such a constructed context may help us gain a better understanding of authentic communication. This study investigates the eect of systematic modi®cation to DCT situational prompts used to elicit requests on the responses of native and non-native speakers of English. Modi®cation included enhancing the situational prompts by adding information on a number of social and contextual variables considered relevant to this study. Results indicate that enhancement did not aect the request strategy or amount of internal modi®cation. However, enhancing situational prompts did produce signi®cantly longer, more elaborated requests in both groups. These ®ndings point to the importance of external modi®cation of speech act production and the need for an instrument that can satisfactorily account for both variation and context. This study has implications for both teaching and testing in interlanguage pragmatics.
DISCOURSE COMPLETION TEST Adapted in 1982 by Blum-Kulka (1982) for the purpose of investigating speech acts, the discourse completion test (DCT) is a questionnaire containing a set of brie¯y described situations designed to elicit a particular speech act. Subjects read the situations and respond in writing to a prompt. An example of a typical DCT prompt is seen below. You missed class and need to borrow a friend's notes. What would you say? (Rose 1992)
518 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
Advantages of using this instrument are well known. Discourse completion tests appear to surpass all other instruments in ease of use. Beebe and Cummings (1996) conclude that DCTs allow researchers to collect a large corpus of data on a wide range of dicult-to-observe linguistic phenomena in a short period of time. More importantly, they note, data elicited with DCTs are consistent with naturally occurring data, at least in the main patterns and formulas. These factors have led to the widespread use of DCTs in sociolinguistic research, including the most extensive research project on speech acts to date, the Cross-cultural Speech Act Realization Project, CCSARP (Blum-Kulka et al. 1989), which investigated requests and apologies across thirteen languages. Numerous other speech act studies have been conducted using this instrument including: apologies (Olshtain and Cohen 1983, Kasper 1989, House 1989a, Linnell et al. 1992, Bergman and Kasper 1993); expressions of gratitude (Eisenstein and Bodman 1986, 1993); refusals (Beebe et al. 1990, Bardovi-Harlig and Hartford 1991); advice (Hinkel 1997); and requests (House and Kasper 1987, Blum-Kulka and House 1989, Faerch and Kasper 1989, de Kadt 1992). Notwithstanding its appeal, the DCT has been the subject of much criticism and a number of validation studies, some of which have found disturbing discrepancies between DCT and spoken data. Beebe and Cummings (1996) found that DCTs failed to elicit the full range of formulas found in spoken data, and that they elicited responses more limited in length and de®cient in the level of elaboration and frequency of repetition typical of human spoken interaction. In their study comparing NS and NNS responses in oral role-plays and in DCT questionnaires, Rintell and Mitchell (1989) found that NS and NNS responses exhibited similar patterns on both instruments although NNS responses were considerably shorter. Kasper and Dahl (1991) view the DCT as in the same methodological category as oral role-plays, namely that of a highly constrained instrument. Although both elicit productive responses, they are both regarded as limited in the authenticity of the situations they represent. Other validation studies have examined instrument-induced variation, comparing data produced in responses to DCTs and to other types of elicitation. Two studies compared DCTs to multiple choice questionnaires. Rose and Ono (1995) found signi®cant dierences between the responses of Japanese speakers in the multiple-choice instrument and the DCT. The respondents avoided or hedged when making requests in the multiple choice questionnaire more than they did in the DCTs, much in the same way Japanese have been observed to behave in real face-to-face interactions. In fact, Rose (1994) came to the conclusion that DCTs may not be the best research instrument to investigate pragmatic behaviour in non-western cultures because of a variety of factors. Hinkel (1997), in her study of English and Chinese advice giving, found that DCTs aected behaviour similarly to the way they did for Rose with requests. She found that on a DCT English speakers chose to give more forms of advice than the Chinese speakers did,
KRISTINE BILLMYER AND MANKA VARGHESE
519
whereas on multiple choice questionnaires the reverse was true. The DCT ®nding runs contrary to socio-pragmatic research which has observed that giving advice is more prevalent in Chinese-speaking societies than in Englishspeaking societies. There have also been eorts to investigate output by modifying the prompt. Most of these studies have focused on the inclusion or exclusion of the hearer response (Blum-Kulka et al. 1989, Rose 1992). Rose (1992) compared data elicited by situations with and without a hearer response added after the situation and concluded that appending a hearer response had no signi®cant eect on the data elicited. Bardovi-Harlig and Hartford (1993), by contrast, found that non-native speakers responded most like native speakers when hearer response was included in the DCT for speech acts requiring a reaction such as rejections. Johnston, Kasper, and Ross (1998) compared three types of hearer responses (positive, negative, and no response) to a variety of speech acts and also found that type of rejoinder had an eect on the choice of strategy. Long before the most recent research comparing the validity of DCT data to other natural and elicited data, Beebe and Cummings (1996) and Wolfson et al. (1989) targeted the design of the instrument as a serious problem. They cited insucient social and situational information in the prompt, omitting such things as background to the event, information on the role relationship between the subject and the imaginary interlocutor, the frequency of their interaction, and details related to context and setting. What is missing according to Beebe and Cummings (1996) is the entire psychosocial dimension, which they point out, sets up a desire on the part of interlocutors to contribute to the interaction in order to maintain their relationship. Without setting the `scene' in a Hymesian sense (Hymes 1972), respondents are left to their own devices, either to invent their own background to the DCT situations, which could vary considerably from respondent to respondent or, more likely, not to invent one at all. By contrast, in natural conversation speakers have access to this very powerful combination of interpersonal and contextual detail, and their unconscious continuous assessment of this information is bound to have an impact on their utterances.
THE ROLE OF CONTEXT AND VARIABILITY IN SLA In the area of pragmatics, Kasper and Dahl (1991) have observed that researchers are dealing with a double layer of variability: the ®rst being that of sociolinguistic variability, and the second being that of variability induced by dierent data collection instruments. In 1966, Labov detected variability among the same subjects depending on the instruments used to collect data, and in a later study how much attention the speaker was paying to what he was saying (Labov 1970). Similar observations were made in other learner language research since then such as Tarone's studies of style shifting (1983, 1985). Ellis (1994) noted variability in learner language in accordance with
520 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
linguistic in¯uences such as the various phonological, grammatical, and syntactical environments, as well as sociolinguistic and psycholinguistic contextual in¯uences. Other studies of variation due to social or task-induced motives have produced unexpected ®ndings. Tarone (1985), for example, found that learners were most target-like in their production of articles and direct object pronouns when they were required to pay the least attention to form, that is, a narrative task. She hypothesized that the narrative led learners to pay most attention to discourse markers such as articles and direct object pronouns, making them produce these forms more correctly. A form±function study that looked at the eect of situation and the relationship between identity and non-target like structures was conducted by Rampton (1987). In his study he observed that learners who had acquired the target language forms purposefully produced some beginner interlanguage forms for `impression management' in order to make fun of their own learner status. Other task-induced variation studies rather than showing the eect of situational context have led researchers to look at psycholinguistic context, in particular, the eect of planning time and focus of attention. Young (1988, 1991) took a multi-factor approach to variability and found that Chinese learners' production of the plural s was dierentially in¯uenced by the situation, the linguistic context, and the redundancy of the plural s, depending on the learner's pro®ciency. Studies on learner variability have also contributed to theories on second language acquisition, such as Ellis's Variable Competence Model (1994) which includes the psycholinguistics constructs of `planning' and `attention' and Preston's Sociolinguistic Model (1989) which comprises such constructs, as well as a social dimension. These studies on inter-learner and intra-learner variability depending on dierent types of contextual in¯uences, be they task-induced, psycholinguistically in¯uenced and/or aected by social factors, point to the enormous importance of context-sensitivity. Given the sensitivity of respondents to the presence, absence, and degree of contextual features and intra-psychic factors, it is even harder to interpret data that may be aected by yet another level of variability, that which the instrument itself produces. Using instruments such as DCTs which arti®cially elicit production, makes it dicult to tell which factors are actually aecting respondents in their written productions. In the case of the DCT the task type has the potential to be very in¯uential. Consider, for example, the eect of additional planning time, which is characteristic of all written production instruments. These problems and those discussed above concerning the construct validity of data collection instruments for pragmatics research raise many doubts about using the DCT to collect sociolinguistic data and might even call into question the value of a study such as this one. However, the advantages of the DCT in relation to its relative speed and facility in administration make it a valuable and still necessary tool in cross-linguistic research and language testing. Furthermore, research which looks at communication within constructed contexts, such as DCTs, helps to improve our understanding of
KRISTINE BILLMYER AND MANKA VARGHESE
521
the constructs involved in authentic communication. For these reasons it is important to explore the capability of constructed instruments to know as fully as possible what their limitations are and how they can best be designed to improve the quality of the data and ultimately the quality of the research they produce. The purpose of the present study is to investigate the eects of enhancing the DCT situational prompt on native and non-native English speakers' responses. In particular we were interested in ®nding out if enriched contextual information, similar to that available to speakers in face-to-face spontaneous interactions, would stimulate more elaboration in support of respondents' communicative goals. This is a feature DCTs have traditionally been unable to encourage. If greater elaboration were the result, such a ®nding might enhance the usefulness of DCTs and similar production questionnaires and bring about a reassessment of instrument design. Since previous research has shown that native speakers and non-native speakers1 react dierently to dierent stimulus types, we felt that it was important to look at the eect of enhanced material on these two groups. This study diers from other DCT design studies which have typically manipulated the rejoinder in that the present study manipulates (and enriches) the internal content of the DCT situation and asks what impact, if any, such manipulation has on the respondents' output. Three research questions guided the present study. 1 2 3
What eect does enriching the content of the situational prompt have on native speaker (NS) output? What eect does enriching the content of the situational prompt have on non-native speaker output? Do native speakers and non-native speakers respond dierently to enriched-content situational prompts?
THE STUDY Construction of the DCT This study examines two versions of a DCT designed to elicit requests. Situations for both versions were derived from descriptions of situations used in the CCSARP project on requests (Blum-Kulka et al. 1989) and later formulated by Rose (1992), but without hearer responses. We decided to use these previously-tested situations for the present study in order to take advantage of the opportunity to compare results with earlier studies and because of the likelihood that the participants in this study would be familiar with the situational prompts. Requests were chosen because they represented an area of pragmatics that the non-native speakers in this study would have the needed pro®ciency to perform on the DCT and would also have occasion to use outside the classroom. Version 1 of the DCT uses the Rose (1992) situations, all of which embed in the situational prompt information on requestive goal, social distance, and
522 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
social dominance. These situations were modi®ed slightly in the following ways. First, to encourage as full a response as possible, the response space was lengthened to 8.5 inches by 4 inches. Next in order to ascertain the respondents' assessment of both level of imposition and the interlocutors' likelihood of compliance, two questions were included after the response space. Finally, the emphasis of the speaker in the two hearer-dominant situations was changed so that subjects would actually take the roles of the librarian and professor rather than try to imagine from a distance how these individuals would respond. Below is an example of a Version 1 situation. An example of how the following situation appeared in the questionnaire for the respondents is provided in Appendix A. Example 1: Version 1ÐMusic Situation You are trying to study in your room and you hear loud music coming from another student's room. You don't know the student, but you decide to ask them to turn the music down. What would you say?
Sociolinguistics and ethnography of communication hold that social context is an important in¯uence on language and language use. Although never empirically tested with production questionnaires per se, countless studies have proven this point very clearly. Fasold (1990) claims that the distribution of linguistic forms is almost totally directed by social factors. Version 2 of the DCT was constructed by examining the literature to identify the type of social and contextual information which critics of DCTs found lacking in the situations and others in the ®eld regarded as necessary and relevant (Wolfson et al. 1989, Beebe and Cummings 1996, Hymes 1972). Hymes' `components of speech' (1972: 59±65) guided our design of the social factors included in Version 2 prompts. These components include a description of the physical setting, including time and place; the participants, meaning interlocutors as well as audience; and the purpose of the speaker, speci®cally the requestive goal. Enhanced prompts speci®ed these social factors by including the following information: the gender and name of the interlocutor, the role relationship, and, by implication, social distance and social dominance, the length of the acquaintanceship, the frequency of the interaction, whether or not the relationship was optional, and a description of the place the interaction happened and the time of day. The interlocutor's emotional or psychological state (e.g. `You felt frustrated/angry/upset/anxious, etc.) was not stated directly in Version 2 prompts as we felt when designing the instrument that other information implied frame of mind. A list of the variables accounted for in each situation is found in Appendix B. We hoped that oering a fuller account of social and contextual data would broaden the opportunity subjects had to more fully understand and interpret the situation without burdening them with excessive length and detail. We were very interested in knowing whether providing the kind of social and contextual information considered crucial to sociolinguistic theory would result in dierences between Version 1 and Version 2 requests. A positive
KRISTINE BILLMYER AND MANKA VARGHESE
523
®nding would add to existing evidence of context-induced variability in datacollection instruments. If dierences were found, it would be important to determine in what ways the collected material was context-sensitive. The few DCT-natural data comparability studies that do exist hold bare-bones prompts at least partly responsible for responses which fail to utilize the range of pragmatic and linguistic resources found in spontaneous communication. Would context-rich prompts be powerful enough to compensate for that eect? We hypothesized that content which provided time and place, more familiarity with the situation, the interlocutors, and the background to the event would be more likely to elicit a fuller range of request strategies, and longer more elaborately modi®ed requests, with a greater degree of supportive mitigation. Example 2 shows the Version 2 Music situation. Here time and place are described, the interlocutor is given a name, and events leading up to the request are included, thus providing the speaker with motivation for the ensuing act. An example of how the following situation appeared in the questionnaire for the respondents is provided in Appendix C. Example 2: Version 2ÐMusic Situation It is 10:30 p.m. on a Wednesday night and you have a paper due the next day. You are trying to ®nish the paper and you can't concentrate because you hear loud music coming from another student's room down the hall. You decide to ask her to turn the music down. The music has been on at this volume for half an hour. You have occasionally seen the student, Lucy Row, in the same dorm during the past six months. She is a student like you, but you have never spoken to her. You have heard other people in the dorm complain about the volume of her music on several occasions although you never have because you study in the library. However, today the library closed early. You are only halfway through and you know that the professor for this class is very strict and does not give extensions. What would you say?
The other ®ve situations used to elicit requests include borrowing a friend's notes, getting a ride, getting an extension on a paper, talking too loudly in the library, and presenting a paper earlier than scheduled. The six situations provide for the usual combinations of social distance and social dominance. The texts of all Version 1 and 2 situations are in Appendix D.
Subjects This study examined the eect of modi®cation to DCT prompts on two groups of speakers. The ®rst group of subjects were thirty-nine native speakers of American English (NSs), who were graduate and undergraduate students at the University of Pennsylvania. Version 1 was administered to twenty students, ten males and ten females. Version 2 was administered to nineteen students, ten males and nine females. The second group of subjects were
524 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
forty-nine non-native speakers of English (NNSs) studying at the English Language Programs of the University of Pennsylvania. These subjects were enrolled in ESL classes at the mid-intermediate level of pro®ciency. Twentyfour were male and twenty-®ve were female. The native languages of these students were Korean (n = 26), Japanese (n = 6), Turkish (n = 4), French (n = 4), Chinese (n = 2), Polish (n = 1), and Kazakh (n = 1). Version 1 was administered to twenty-three students. Version 2 was administered to twentysix students. Data were collected primarily in classrooms, and subjects were told that the data were being collected for a research project in the Graduate School of Education. The exact purpose of the study was not disclosed. Subjects were not given a time limit to complete the questionnaires, and they were instructed not to use dictionaries.
Data coding and analysis Data were coded using the Blum-Kulka et al. (1989) coding scheme, which provides detailed descriptions of request head act strategies, and forms of internal and external modi®cation. Detailed examples will be provided in the results. The main categories of analysis were as follows.
Level of directness of request strategy head acts The ®rst category of analysis for NSs and NNSs comprised the classi®cation of request head acts according to level of directness. A head act is de®ned by Blum-Kulka et al. (1989) as `the minimal unit which can realize a request'(p. 275) and excludes those parts of the act sequence which are not essential. Blum-Kulka et al. (1989) consider directness a pragmalinguistic category. It describes the degree to which speaker intent is apparent from the locution. We began with the CCSARP (Blum-Kulka et al. 1989) project's coding scheme and coded request strategies into nine dierent types as shown in Appendix E. We then followed Brown and Levinson's (1987) classi®cation of head act strategies by collapsing these strategies into the following three categories: 1
2
Direct strategies (D), where the speaker's intent is apparent from the locution. This transparency is derivable from the grammatical mood (imperative), the use of performative verbs, or the semantic content of the utterance, such as: Please turn the music down. Conventionally indirect strategies (C), where the interpretation is aided by conventional usage, such as (a) a suggestory formula or (b) reference to a preparatory condition, as in: How about turning the music down? Could you turn the music down, please?
KRISTINE BILLMYER AND MANKA VARGHESE
3
525
Non-conventionally indirect strategies (N), which includes strong or mild hints, such as: It's very noisy so late in the evening.
Internal modi®cation of the head act Internal modi®cation of the head act comprises syntactic and lexical or phrasal modi®ers within the head act itself, which serve to either soften or intensify the force of the request. The data consisted primarily of lexical or phrasal downgraders, which soften the request and a few instances of upgraders which intensify the request. Lexical downgraders, include politeness markers (please), consultative devices (do you think), understaters or hedges (Could you do your paper a bit earlier?). Syntactic downgraders include the use of tense and aspect (I was wondering if . . . ), conditional clauses, and the use of the interrogative.
External modi®cation of the head act Supportive moves. Speakers can mitigate the face-threatening forces of acts like requests with moves that prepare the hearer for the upcoming request, oer reasons or explanations for the request, or remove potential objections to the request. These are called supportive mitigating moves, and it is a category of considerable interest in this study because it is often the part of the request act that DCTs fail to elicit and where dierences between NS and NNS performance are often seen. Supportive moves were coded following BlumKulka et al., 1989 for the following types of mitigation: preparator (Can I ask you a question?), getting a precommitment (Could you do me a favour?), disarmer (I'll give your notes right back), grounder (I had trouble with the data collection), imposition minimizer (If you have a few minutes, can I talk with you?), and promise of reward (You can borrow my notes anytime). We coded supporting mitigating moves separately because we wanted to know if we could ®nd a relationship between context supplied in the two versions of the DCTs and variety of mitigation. We also wanted to gain a fuller understanding of the composition of requests with multiple supporting moves. Aggravating moves included insults, threats, and moralizing. Alerters. Alerters serve to warn the hearer of an upcoming speech act and include names and address terms, such as Tom or Professor Smith or attention getters such as Excuse me. Requests were coded for the presence or absence and not quantity of alerters per request, and so were counted as one for an address term, attention getter, or a combination of both.
Length of the entire request act The entire request act is comprised of the head act, including all internal modi®cation, and all external modi®cation which came either before or after the head act. We did not include any parenthetical remarks made by
526 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
respondents or reported speech forms. Each word in the request act was counted.
Statistical analysis Statistical analyses were conducted on data situation by situation to determine whether dierences between versions were signi®cant. Tests were performed on NS and NNS data independently. These data are presented in Tables 1±6. For purposes of discussion and clari®cation a tally of situational ®ndings for each measure was included in each table but not tested for signi®cance. This was not done since previous studies stress the importance of keeping situational ®ndings independent. Combining them can eace the eects of situational variability, which we consider relevant to this study. Chi-square tests2 were performed in two instances: level of directness of request strategy and frequency of requests containing alerters. Chi-square tests are used to determine whether or not there is an association between categorical measures. In this study only these two measures met that criterion. Since ttests are used to compare the means of quantitative measures, we used twotailed t-tests to establish whether the means for the internal modi®cation, supportive moves, and length were signi®cantly dierent between Version 1 and Version 2 data. The same tests of signi®cance were performed on the NS and the NNS data independently. Discussions of similarities and dierences between the NS and NNS ®ndings were based on these independently performed tests. Statistical comparisons of native and non-native measures of performance were not conducted as we felt that the independent tests of signi®cance were of sucient explanatory value to stand on their own.
RESULTS AND DISCUSSION Level of directness of request strategies Native English speakers A chi-square analysis of the NS data by situation reveals statistically signi®cant dierences in only one of the six situations, the Music situation (Table 1). Native speakers responded with virtually the same level of directness to both context-impoverished and context-enhanced prompts in the other ®ve situations. NS respondents to both DCT versions used conventionally indirect request strategies in over 80 per cent of their requests (Table 2). Although not statistically tested, dierences were noted in the second and third ranked strategy types. After conventionally indirect strategies Version 1 respondents preferred direct strategies followed by non-conventionally indirect strategies. Version 2 respondents chose non-conventionally indirect strategies second and direct strategies last. Overall, Version 2 respondents used proportionately more indirect strategies (of both kinds) than Version 1 respondents did.
KRISTINE BILLMYER AND MANKA VARGHESE
527
Table 1: Level of directness of NS and NNS request strategies by situation Version Native speaker 1 (n = 20) 2 (n = 19) p < 0.05 Non-native speaker 1 (n = 23) 2 (n = 26) p < 0.05
MUSIC D C N
NOTES D C N
RIDE D C N
LIBRARY D C N
EXTEN. D C N
PRESEN. D C N
4 15 0 2 18 0 0 19 0 0 18 1 p < 0.034* p < 0.226
0 13 1 2 16 1 p < 0.452
8 9 1 2 14 1 p < 0.097
0 14 2 1 10 8 p < 0.08
1 14 1 1 16 0 p < 0.90
6 14 2 8 16 0 p < 0.311
5 15 1 4 21 0 p < 0.411
9 11 2 15 11 0 p < 0.175
3 17 2 6 17 1 p < 0.536
12 7 1 9 11 0 p < 0.314
3 18 0 3 17 0 p < 0.948
Level of directness: Direct (D), Conventionally indirect (C), Non-conventionally indirect (N)
The Library (n = 8) and Music (n = 4) situations appear to account for most of the direct requests (D) made by Version 1 respondents. It is interesting to note that in both situations the respondent was the victim of an objectionable level of noise made by the hearer. It is also interesting to note that even though respondents were well within their rights to ask the hearer to refrain from this oensive behaviour, Version 2 respondents chose to cushion the request by making predominantly conventionally indirect (C) requests. It is possible that Version 2 prompts better personalized the oending interlocutor in the Music and Library situations resulting in less direct Version 2 requests. The same principle may be called upon to account for the large number of non-conventionally indirect (N) requests found in the Version 2 Extension
Table 2: NS and NNS level of directness in request strategies for combined situations Directness
Native speaker Direct Conventionally indirect Non-conventionally indirect Non-native speaker Direct Conventionally indirect Non-conventionally indirect
Combined Version 1 situations
Combined Version 2 situations
Total versions and situations
14.5% (n = 15) 80.6% (n = 83) 4.9% (n = 5)
5.5% (n = 6) 84.5% (n = 93) 10% (n = 11)
9.9% (n = 21) 82.6% (n = 176) 7.5% (n = 16)
29.7% (n = 38) 64.1% (n = 82) 6.2 (n = 8)
32.4% (n = 45) 66.9% (n = 93) 0.07% (n = 1)
31.1% (n = 83) 65.5% (n = 175) 3.4% (n = 9)
528 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
situation (n = 8). The added contextual detail in the Version 2 prompt may have made respondents feel an extra measure of awkwardness about asking a professor for an extension, thus causing almost half of the respondents to choose a hint.
Non-native English speakers As Table 1 shows, the chi-square analysis of the NNS data revealed no signi®cant dierences between Version 1 and Version 2 responses for any of the six situations. In other words the level of directness of request act strategies for NNSs does not appear to be sensitive to the type and amount of information in the situational prompt. A closer look at the distribution of request strategies by level of directness for all situations combined shows hardly any variation between versions. Although this comparison was not statistically tested, respondents used conventionally indirect strategies for 64 per cent of the Version 1 and 67 per cent of the Version 2 requests (Table 2). The order of preference for level of directness was the same for respondents to both versions: conventionally indirect, followed by direct, and followed by non-conventionally indirect.
Comparison of native English speakers and non-native English speakers ®ndings Independent statistical analyses con®rm for both NS and NNS populations that level of directness does not appear to be sensitive to prompt construction. The data also show a very high incidence of conventionally indirect requests for both NSs and NNSs. This indeed has been the main ®nding of the CCSARP project as well as the main ®nding of request studies using naturally occurring data. In fact, the area of main patterns and formulas is the most widely cited category of analysis where typical DCT data seem to replicate spoken data. However, there may be some dierences between the NS and NNS choice of directness which are easier to see by the compiled situational data in Table 2. These data are not statistically tested, so our comments regarding dierences here are speculative. The comparison shows that conventionally indirect strategies accounted for 82.6 per cent of the NS requests; whereas, they accounted for 65.5 per cent of the NNS requests. Furthermore, NNSs used direct strategies three times more frequently than did NSs. This dierence was particularly evident where NNS subjects were asked to take on the role of a higher status speaker, as in the Library and Presentation situations (Table 4). In both situations the NNSs' status as an authority appeared to sanction their use of more direct requests while the same shift in role to higher status interlocutor did not appear to have such an eect on NSs. Hoppe-Gra, Herrmann, Winterho-Spurk, and Mangold (1985) and House (1989b) provide a possible explanation for the NNS ®nding in the concept of the standard situation. Situations like the Library and Presentation situations can
KRISTINE BILLMYER AND MANKA VARGHESE
529
be considered standard because they are situations where speakers' and hearers' rights and obligations are relatively set in terms of the parameters that regulate interactional work, making the likelihood of negotiation quite low because speakers know both `where and who they are' (House 1989b: 107). So, in the case of NNSs in this study, the addition of contextual information did not appear to alter their assessment of their roles as higher status interlocutors or their consequent right to make a direct request of someone obligated to comply with it. Although we did not analyse the nonnative speaker data according to the respondents' ®rst language, it may have been useful to do so here as language transfer is quite likely to have played a role in this result. Future examination of this data should focus on such an analysis.
Internal modi®cation of the head act Native English speakers A two-tailed t-test on the mean downgraders per request revealed signi®cant dierences between responses to Version 1 and 2 prompts only in the Notes situation (Table 3). In fact the low mean value of downgraders across all situations in all versions suggests that NS subjects are not mitigating their requests in the head act very much at all.
Table 3: Internal modi®cation: NS and NNS mean downgraders per request Version Native speaker 1 (n = 20) 2 (n = 19) p < 0.05 Non-native speaker 1 (n = 23) 2 (n = 26) p < 0.05
MUSIC
NOTES
1.1 1.3 p < 0.133
0.4 0.7 1.0 0.7 p < 0.000* p < 0.711
0.8 0.5 1.3 0.9 p < 0.015* p < 0.131
RIDE
0.5 0.6 p < 0.564
LIBRAR.
EXTEN.
PRESEN.
0.8 0.9 p < 0.733
0.5 0.6 p < 0.072
0.5 0.6 p < 0.972
0.8 0.9 p < 0.741
0.2 0.5 0.5 0.7 p < 0.015* p < 0.380
Non-native English speakers Results of a two-tailed t-test on the NNS data showed that enhancing the content of the situation had very little impact on internal modi®cation (Table 3). Only two of the six Version 2 situations (Music and Extension) resulted in signi®cantly greater use of internal modi®cation.
530 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
Comparison of native English speakers and non-native English speakers ®ndings In the category of internal modi®cation the NS and NNS ®ndings were similar. The use of downgraders was only minimally sensitive to material in the prompt. A closer examination of the mean number of downgraders per request for each population shows that, with only a few exceptions, NSs and NNSs internally modi®ed the head act less than once per response. This pattern held across versions and situations for both populations. Once again, these ®ndings are similar to those of the CCSARP study. Thus far, our analysis of the core of the request, the head act, shows two things: ®rst, that manipulation of situational content has no signi®cant eect on choice of request strategy according to level of directness or amount of internal modi®cation; and second, that the preference for conventionally indirect strategies seen here across versions and populations is consistent with previous studies which examine both naturally occurring and DCT data.
External modi®cation of the head act: supportive moves Native-English speaker ®ndings Table 4 gives the results of a two-tailed t-test of mean supportive moves per request for NSs and shows signi®cant dierences in four of the six situations. By examining these four situations one can see that the mean number of supportive moves in data elicited by the elaborated Version 2 situations was two to four times greater than the mean number of supportive moves elicited by the context-poor Version 1 situations. The mean number of supportive moves across all Version 1 data was 1.0 per request. By contrast, respondents to Version 2 situations oered an average of 2.2 supportive moves per request, more than double the number of mitigating moves. Data from the Presentation situations illustrates the dierences between responses elicited by Version 1 and Version 2 prompts. I was really hoping that you could present your paper one week earlier. (Version 1: Presentation)
This request contains no external modi®cation in the form of supportive moves. The above example can be contrasted with a typical example of a request act elicited by the Version 2 Presentation situation, where a number of supportive mitigating moves are evident. Nancy, you are one of the strongest students in the department so I am hoping you can do me a favor. If you can't, it's no problem but we're studying the subject relevant to your presentation then. Can you get it ready? If not, it's okay. (Version 2: Presentation)
The above response elicited by the elaborated prompt in Version 2 contains two imposition minimizers (If you can't and if not, it's okay), a grounder (we're
Table 4: External modi®cation: NS and NNS mean supportive moves per request Version
NOTES
RIDE
LIBR.
EXTEN.
PRESEN.
Combined
1.4 1.9 p < 0.765
1.2 2.4 p < 0.014*
1.1 2.1 p < 0.024*
0.6 1.2 p < 0.059
1.3 3.1 p < 0.000*
0.5 2.3 p < 0.000*
1.0 2.2
1.8 2.9 p < 0.020*
2.0 3.3 p < 0.082
3.2 4.1 p < 0.243
1.1 2.0 p < 0.037*
2.0 3.6 p < 0.004*
1.5 3.5 p < 0.000*
1.9 3.2
KRISTINE BILLMYER AND MANKA VARGHESE
Native speaker 1 (n = 20) 2 (n = 19) p < 0.05 Non-native speaker 1 (n = 23) 2 (n = 26) p < 0.05
MUSIC
531
532 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
studying the subject relevant to your presentation) and a precommitment (I am hoping that you can do me a favor). When we compared the types of supportive moves elicited by both versions, we found that Version 2 data contained many more promises of reward (I'll be more lenient with you for the grading) and disarmers (I know you have a lot of work) than did Version 1. Version 1 data consisted primarily of reasons or excuses for the request. Version 2 respondents also oered other speech acts, such as compliments (as seen in the example above), apologies, expressions of gratitude, and other solidarity and face-saving moves. These hardly ever appeared in the original Version 1 data. The amount of mitigation in the request act did not seem to be primarily a factor of social dominance. If it were, we might have seen a larger number of supportive moves made by respondents who took the part of lower status speakers in the Extension and Ride situations. Although the greatest number of supportive moves per request was found in the Extension data, a large number of mitigating moves were made in other situations. For example, in the Presentation situation, the speaker (who takes the role of the professor) is dominant while in the Notes situation both the speaker and hearer are status equals. Wolfson's Bulge Theory (1989) could account for the increase in mitigation between status equals, in that individuals in the centre of the social distance continuum are more likely to engage in longer interactions with more extensive elaboration. Even so, this theory does not account for the increased mitigation in the Presentation situation, where social distance is greater, or for the fact that in both of these situations we saw greater use of supportive moves in the enriched Version 2 but not in the original Version 1.
Non-native English speaker ®ndings A two-tailed t-test of the NNS data revealed that enhancing the content of the situations resulted in a signi®cantly greater use of supportive moves in responses to four out of six Version 2 prompts (Table 4). Overall the mean number of supportive moves elicited by Version 1 situations combined was about 1.9 per request. The mean for Version 2 prompts was 3.2 moves per request. The dierences between typical Version 1 and Version 2 non-native responses are similar to the dierences found in the native speaker data, as illustrated by the examples from both versions below: I'd like to change our schedule, so why don't you present a paper a week earlier than scheduled. (Version 1: Presentation)
This Version 1 Presentation request begins with one supportive move, (I'd like to change our schedule), in this case a grounder, which gives reason for the request. This request can be compared with a typical request elicited by the enriched Version 2 Presentation prompt:
KRISTINE BILLMYER AND MANKA VARGHESE
533
Nancy, I know you have a heavy course load because midterm exams are next week. However, I want you to present a paper in a class next week. Maybe it is dicult but I know you are really an excellent student. Can you do that for the class? If you can do that, I'll give you some bonus score. (Version 2: Presentation)
This request contains three supportive moves: two disarmers (I know you have a heavy course load because midterm exams are next week and Maybe it is dicult), and one promise of reward (I'll give you some bonus score). It also contains one compliment (I know you are an excellent student).
Comparison of native English speakers and non-native English speakers ®ndings As the independently conducted statistical analyses show both NS and NNS respondents exhibited similar patterns in supportive moves. Version 2 prompts elicited signi®cantly more supportive mitigating moves than Version 1 prompts for most situations for both populations (Table 4, p < .05). In general, Version 2 requests were characterized by a greater and more diversi®ed set of supportive moves than Version 1 requests. Beebe and Cummings' (1996) observations help to provide an explanation for these dierences. Typical DCT prompts, similar to Version 1 prompts, do not, `bring out the real psychological (or perhaps more accurately, `psychosocial') dynamics of a natural interaction between members of a group: the establishment or maintenance of one's own reputation and of rapport with the human being one is addressing' (Beebe and Cummings 1996: 8). The reason may be that respondents are addressing an anonymous, characterless entity and have no motivation to establish or preserve a relationship. The original Version 1 prompts tended to elicit minimalist response data. By contrast, Beebe and Cummings maintain, in real life situations, people repeat, hedge, and excuse themselves more because they are interacting with a real person. We argue that enhanced social and situational content in Version 2 prompts may have furnished respondents with more information to establish or maintain their reputation and rapport with the addressee. As the data on supportive moves suggest, when the social dimension of the situational prompt is augmented, then the written responses become more elaborate in much the same way as speech in natural spontaneous interaction: with excuses and reasons, promises, and other means of saving one's own face and minimizing the potential damage to another's. Missing from our Version 2 prompts was information on the interlocutors' inner states. Respondents were left to infer (from other contextual information) or invent their own and their interlocutor's personality and psychological state. Had we provided more explicit information of this type in Version 2, we may have seen additional dierences in the area of supportive moves. Although in most ways NS and NNS ®ndings were very similar, there were
534 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
some interesting dierences in the overall amount of external support. The combined tallies of all situations shown in Table 4 indicate that the NNSs in general used more supportive moves in response to both Version 1 and Version 2 prompts than did the NSs. These average scores were not tested for signi®cance. Interestingly, however, the overall dierence in the amount of external modi®cation between Version 1 and Version 2 responses was almost the same for native English speakers (1.2 moves) and for NNSs (1.3 moves).
External modi®cation of the head act: alerters Native-English speaker ®ndings The results of a chi-square test comparing frequency of requests containing alerters showed signi®cant dierences between Version 1 and Version 2 data in three situations (Table 5). Overall we found that almost three times as many Version 2 requests contained alerters as Version 1 requests. Twenty-®ve percent of Version 1 NS requests contained one or more alerters while 61 per cent of the Version 2 NS requests had one or more alerters per request.
Non-native English speaker ®ndings NNS use of alerters showed no clear pattern of use. As Table 5 illustrates, there were statistically signi®cant dierences between versions in three of the six situations. However, only one of these situations (Presentation) gave us results in the expected direction: a greater frequency of requests with alerters in Version 2. Ride and Library requests showed just the opposite pattern: there were more requests containing alerters in the Version 1 data than in the Version 2 data. The combined situational data for NNSs shows almost the same total number of requests containing alerters in response to each version (V1 = 74, V2 = 79).
Comparison of native English speakers and non-native English speakers ®ndings The combined situational alerter data were not statistically tested, but for purposes of easier comparison they are presented in Table 5. For NSs more Version 2 requests contained alerters than did Version 1 requests. This was not the case for the NNSs. Overall there was virtually no dierence between NNS Version 1 and 2 data on this measure. We do not know why the pattern of use between the NS and NNS populations was dierent. It is easy to attribute the increased frequency of alerters in the NS population to the fact that ®ve of the Version 2 prompts supplied the interlocutor's name. That information along with the additional personal and situational information may have succeeded in prompting the respondent to frame the request with a personalized alerter. We think it likely that the same held true for the Library situation, even though no names were
Table 5 NS and NNS frequency distribution for alerters Version
6
NOTES 3
RIDE 10
LIBR. 3
EXTEN 3
PRESEN. 1
Combined 26
10 p < 0.189
12 p < 0.002*
14 p < 0.886
8 p < 0.053
12 p < 0.008*
11 p < 0.001*
67
13 11 p < 0.369
13 12 p < 0.901
19 16 p < 0.036*
15 10 p < 0.013*
11 16 p < 0.286
3 14 p < 0.000*
74 79
KRISTINE BILLMYER AND MANKA VARGHESE
Native speaker 1 (n = 20) 2 (n = 19) p < 0.05 Non-native speaker 1 (n = 23) 2 (n = 26) p < 0.05
MUSIC
535
536 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
supplied. Personalizing the situation a bit more made it more likely for the librarian to soften the interruption with `Excuse me,' the typical alerter which framed most NS Library responses. An explanation for the NNS pattern of use is far more dicult to construct. When looking at the raw data, we found that NSs made more use of the names supplied in the prompts than did NNSs. One possible explanation for this ®nding is cultural dierences in rules governing forms of address. Again, since we did not select non-native speakers from the same ®rst language, it would be dicult to assess how language transfer played a role here. However, the value of contrastive pragmatics is fundamental to this type of research, and future validation of this instrument should include it.
Discussion of external modi®cation In the overall category of external modi®cation, one could argue that the information supplied in the enriched Version 2 situations (i.e. names, background to the relationship, contextual details) is arti®cially inducing respondents to do more supportive work and use alerters. This is probably true. On the other hand, real social interactions are always framed within a context, even if only a partially complete one. Interlocutors are cognizant of here-and-now factors. If there is a background to the relationship or a history, participants know it, and they are likely to know the interlocutor's name in such a case. They can also make an educated guess about the likelihood of future interaction with the interlocutor. Such knowledge helps them to oer reasons or excuses more easily, remove a potential objection to the request, cajole, ¯atter, or do whatever seems appropriate given the person, time, and place. In the elaborated Version 2, respondents were given a partially complete picture of a contextualized relationship and the option, as they would have in real life, to use the information provided or not. In fact, in the original Version 1 data, ®ve NS respondents noted in parentheses that they would give a reason for not being able to complete their paper even though they were unable to formulate one at the time in writing. This bears out our belief that the desire to mitigate the head act in some way is present in the respondents' minds, even if not in their actual responses, and an enhanced prompt can help respondents realize that need. The exception to the ®nding of more frequent mitigation in Version 2 requests was in the category of alerters for NNSs. Here, the results were mixed: NNSs constructed requests containing alerting signi®cantly more often in two context-poor Version 1 situations and signi®cantly more often in only one Version 2 situation (p < .05). As mentioned above, this response may have been related to ®rst language pragmatic interference, or it may have been due to the lengthy and challenging nature of the context-rich Version 2 where names may have been `lost' in the information supplied.
KRISTINE BILLMYER AND MANKA VARGHESE
537
Length of entire request act Native English speaker ®ndings A two-tailed t-test revealed signi®cant dierences between versions in this category for ®ve of six situations in the mean length of utterance (MLU). As Table 6 shows, NS utterance length in the enriched Version 2 data ranged from two to three times longer than NS utterance length in the original Version 1 data. The MLU for Version 1 requests across all situations was 16.9; for Version 2 requests it was 34.1, almost a twofold increase. These dierences are illustrated by a typical native speaker request from each version. The ®rst is a response to the Version 1 Music prompt. I'm trying to study. Could you please turn the music down a little? (Version 1: Music)
The utterance length for this request was thirteen, consisting of the head act with one lexical downgrader (a little) and a supportive move, a grounder, which gives a reason for the request (I'm trying to study). The next example is a request in response to the elaborated Version 2 Music prompt. Lucy, I'm really sorry to bother you, but if possible could you please lower the volume a little. Tomorrow I have a paper due and I'm really stressed out. (Version 2: Music)
The utterance length for this request was twenty-six words. This example includes the head act with two lexical downgraders, a name alerter (Lucy), three supportive moves which include a disarmer (I'm really sorry to bother you), and two grounders (I have a paper due and I'm really stressed out). Interestingly the requestive head act in each example is almost identical. The dierences between Version 1 and 2 response data lie almost exclusively outside the head act, and within the supportive moves.
Non-native English speaker ®ndings Enriching the content of the DCT prompt appears to have had a similar eect on the NNS mean length of utterance. As Table 6 shows the length of the entire request act was signi®cantly greater (p < .05) in four out of six Version 2 situations (Music, Notes, Extension, and Presentation). The MLU across all NNS Version 1 situations was 21.8; it was 35 for all Version 2 situations combined. The following is a typical example of a NNS response to the Version 1 Music prompt: Excuse me, could you turn the music down a little? I'm studying now. (Version 1: Music)
The utterance length for this request was thirteen, and it consisted of an alerter (Excuse me), the head act with one lexical downgrader (a little), and one supportive move, a grounder, which oers a reason for the request (I'm
Version Native speaker 1 (n = 20) 2 (n = 19) p < 0.05 Non-native speaker 1 (n = 21) 2 (n = 26) p < 0.05
MUSIC
NOTES
RIDE
LIBR.
EXTEN
PRESEN.
Combined
14.2 28.8 p < 0.001*
12.8 35.3 p < 0.001*
24.4 31.0 p < 0.157
10.0 19.3 p < 0.005*
22.6 47.7 p < 0.000*
17.3 42.3 p < 0.000*
16.9 34.1
18.1 29.7 p < 0.005*
21.7 37.3 p < 0.004*
30.1 39.2 p < 0.074
15.3 19.8 p < 0.152
22.6 42.3 p < 0.000*
23.1 41.5 p < 0.000*
21.8 35.0
538 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
Table 6 NS and NNS mean words per request
KRISTINE BILLMYER AND MANKA VARGHESE
539
studying now). This can be compared with an NNS request elicited by the Version 2 Music prompt: Hi! Do you remember me? I had lived in the same dore during the past six months. And I must hand out my paper tomorrow. Please the volume down. (Version 2: Music)
This example has a length of twenty-nine words, and it includes a greeting (Hi!), three supportive moves, including two preparatory utterances to establish the level of acquaintance (Do you remember me?, and I had lived in the same dore. . .) and a grounder (I must hand out my paper), and the request act itself containing one lexical downgrader (please).
Comparison of native English speakers and non-native English speakers ®ndings Mean length of utterance appears to be sensitive to prompt material for both native and non-native English-speaking populations. The mean length of utterance (MLU) was signi®cantly greater for NSs in ®ve Version 2 situations and for NNSs in four Version 2 situations. This ®nding is not surprising and can probably be attributed to the increase in supportive moves and other speech acts such as greetings, expressions of gratitude, compliments, and other speech acts, which Version 2 prompts appear to elicit more successfully. Though not statistically veri®ed, it is interesting to note, however, that the MLUs were similar, with NNS MLUs being slightly longer in both versions. The combined MLU in Table 6 shows this comparison. Blum-Kulka and Olshtain (1986) reported similar results in a study of utterance length, ®nding that NNS utterance length exceeded NS utterance length when learners were at approximately the intermediate level of pro®ciency. They attributed greater NNS length to the increased number of and overly wordy supportive moves, these speakers not having yet achieved the verbal eciency evident in NS responses. Edmondson and House (1991) reported similar ®ndings, attributing the increase in NNS verbosity to a `waing' phenomenon, a tendency towards being overly informative not evident in native speakers' responses. Sometimes NNSs incorporated parts of the initiating move or stimulus into their response move. Blum-Kulka and Olshtain (1986) suggest that learners' lack of con®dence in their ability to get the message across can result in the use of excessive contextual information and more words.
Other results Once the categorical coding was completed, we returned to look at several other attributes of the data which the coding scheme was not able to capture. Although we discuss characteristics of some native speaker data, we focus primarily on the quality of non-native responses to the prompts.
540 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
Interpersonal quality of Version 2 requests For both the NS and NNS data, where the need to mitigate or express urgency was strong, Version 2 requests appeared to more fully realize this need with a more personalized tenor, speci®c to either or both the respondent and the imagined interlocutor. The content of these requests often included disclosure of more personal information, expression of attitudes or feelings, use of nicknames or terms of endearment, and sometimes a show of vulnerability or an admission of weakness. We found Version 2 request sequences to include preparators which served to establish or reestablish the role relationship, and pre-sequences such as pre-favours, apologies, or other solidarity moves which showed concern for the interlocutor and the future of the relationship. Although the explanatory content of these requests did often come from the prompt, we also noted a great deal of `invented' content, which was not related in any way to the information contained in the prompt. The nature of these elements is illustrated by the following comparison which begins with a Version 1 request by a NNS. Hi. I missed class. Would you mind if you handed your notes? (Version 1: Notes)
This respondent employed an alerter, one supportive mitigating move, and a conventionally indirect request strategy. By contrast in the Version 2 request below the speaker calls the hearer by name, he owns up to his past borrowing history, and he shows a willingness to reciprocate by oering the same kind of help to the hearer in the future. Tom, would you mind if I want to borrow your notes? I know it's the third time that I asked for but I had no chance to join the class last week. I'll be happy to help you if you need my notes in other classes. (Version 2: Notes)
The native speaker data contained similar elements of `personalizing' the request. We also found some instances of creative inventing as the lengthy preamble to this native speaker request shows. Hey, Tom, how's it going? Have you been keeping up with the 76ers? No? well, I tell you last week's game was incredible. Yeah, it went into triple overtime and the 76ers won. Afterwards, my roommates dragged me out to the local bar and we had a few drinks. Unfortunately, I was a bit hung over and missed class. I know I've already borrowed your notes twice this semester but I was wondering if I could see last week's notes. I have an old exam and we should study for the ®nal together next week. I'll give you a call. Hey, take care and go 76ers! (Version 2: Notes)
The background to this speaker's request has nothing to do with the content supplied by the text of the prompt. He appears to have imposed a scenario relevant to his world on top of, but still ®tting well with, the prompt. This
KRISTINE BILLMYER AND MANKA VARGHESE
541
response also captures the nature of the relationship the respondent has in his mind with the interlocutor. What the perlocutionary eect a personalized request has on a hearer is not known. Personalizing a request appears to make it more potent. However, depending on the status of the interlocutor and the role relationship obtaining between participants, personalizing a request in the ways we described above could lead just as easily to pragmatic failure as to success, especially in cases where more distance and deference is required. Even though we did not detect an inappropriate level of friendliness in higher status or greater social distance requests, a more rigorous analysis of this observation would be useful. Overall, however, it appears that prompts which provide a richer interpersonal context help respondents to envision a more complex relationship with their interlocutor and call upon a wider array of cognitive and linguistic resources to carry out the request.
Challenges posed by Version 2 prompts The lengthy nature of the Version 2 prompts may have posed considerable challenges to the NNS respondents in this study. The ®rst challenge was reading and understanding the rather long situational prompt. As mentioned earlier, this may have been one of the reasons NNSs formulated fewer of their Version requests with alerters. NNS respondents took approximately 20 minutes longer to complete Version 2 DCTs than Version 1 DCTs. But they did not make more mistakes in interpreting the situations. In fact, in one case, the longer, more complicated Version 2 Presentation prompt appeared to have helped clarify the addressee of the request; in the Version 1 prompt a number of respondents were confused about whether the addressee was an individual or the whole class. So, in fact, the enhanced content in this case helped disambiguate the prompt. An additional diculty for NNSs was to discriminate between prompt content that was suitable and unsuitable for integrating into the request. It is clear that not all content is equal in the Version 2 prompts, and some would be inappropriate for reference to or inclusion in the request. Nevertheless, NNS respondents seemed to have fared well on this challenge since no inappropriate information from the prompts appeared in the requests.
Use of exact wording Although we fully expected to ®nd propositional content from the enriched prompts in the requests, we were concerned that Version 2 situations would provoke widespread borrowing of exact linguistic material from the prompts, especially by NNSs to use in their requests. While there were some exact word lifts, there were no dierences between Version 1 and Version 2 responses for either population. Respondents to both versions focused on a few set phrases, and incorporated them into their requests (e.g. from the Music situation: can't
542 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
concentrate, paper due; Ride: supposed to, seen you on occasion). However, this borrowing constituted a very small percentage of the NNS responses. Most of the content was the students' own original formulation, and it was representative of their level of pro®ciency.
Encoding of formulaic expressions There was ample evidence in both versions that NNSs did not have mastery of the conventionalized routines most commonly used by native speakers of American English to realize the request head act. In the Library situation, the forms, be quiet or quiet down were often expressed by our NNS respondents by a variety of interlanguage forms, such as speak slowly, do your voice down, make your voice a bit smaller, turn down the voice, or have silence. In the Ride situation NNSs requested a ride by asking, Would you mind if I take your car in the morning?, I'd like to go with you when I come back after school, or I need to ride a car. These interlanguage features were found in both versions, apparently occurring independently of the prompt material. However, Version 2 requests containing incorrectly encoded request expressions were often easier to disambiguate than Version 1 requests with faulty head acts. This was the case because Version 2 requests were often embedded within more supportive material, making the intent of the request more transparent.
SUMMARY OF FINDINGS It appears that certain components of the request act are sensitive to variation in the internal structure of the DCT, but others are not. The major ®ndings were as follows. 1
2
We did not ®nd dierences, signi®cant or meaningful, attributable to the content of the DCT prompts for either population of respondents on either measure of the request head act itself, speci®cally, the distribution of head act request strategies according to level of directness, or the frequency of lexical and syntactic mitigation internal to the request head act. Both native and non-native English speakers used internal modi®cation less than once per request. We did ®nd that adding material to the situational prompt had a signi®cant eect on response data for both native and non-native English speakers on two measures: mean length of the entire request act and amount of external modi®cation. Speci®cally, we found that for most situations, the mean length of entire request act prompted by the elaborated Version 2 situations was two to three times greater than the mean length of the request act prompted by the brief Version 1 for both native and non-native English-speaking respondents. In the category of external modi®cation we found that the mean number of supportive moves was two to three times greater in the elaborated Version 2
KRISTINE BILLMYER AND MANKA VARGHESE
3
543
responses than in the original Version 1 responses for both populations for the majority of situations. In four of the ®ve categories of analysis the response patterns of the native and non-native speaker respondents to Version 1 and Version 2 material converged: level of directness of head act, use of internal modi®cation, external modi®cation of the head act by means of supportive moves, and mean length of utterance. Only in the category of alerters did patterns of use between the two populations studied look dierent.
CONCLUSION Overall, it appears that the content-enriched DCT prompts elicit more robust external modi®cation and elaboration than do the archetypal content-poor prompts which most DCT studies to date have used. Our results also support others' ®ndings that the few social variables that are typically embedded in production questionnaires (social distance, dominance, and imposition) are not powerful enough on their own to elicit as full a response as is found in naturally occurring data (Meier 1995, Holmes 1990, Fraser 1990 as cited in Hinkel 1997). Since critics of DCTs and similar production questionnaires usually point to the failure of these instruments to elicit elaboration, our ®ndings have implications for instrument design and construct validity. The request head act, on the other hand, appears to be quite resilient to instrument-induced variability. This conclusion is borne out by the high incidence of conventionally indirect strategies found in this study regardless of the prompt material, and in previous DCT studies of requests using elicited and observational methods. This ®nding may have important implications for second language teaching and testing. In their survey of ESL textbooks which teach the rules of speaking Billmyer et al. (1989) found that lessons typically focus on the learning of canonical head act formulae used to realize various speech acts, but they rarely attempt to develop learners' ability to preface or follow up a speech act with moves that can support their communicative goals. Prompt material is usually too thin to elicit elaboration, and because ESL textbooks do not treat external modi®cation through aggravation or mitigation explicitly, there is usually no way for them to learn about or practice the many strategies (grounders, preparators, disarmers, minimizers, etc.) available to them. Our ®ndings suggest that prompt material can be constructed in ways that will stimulate elaboration during simulated classroom practice situations. This is very important for language classrooms where supplying learners with context-enriched support can help them envision a wider range of social relationships than those which occur naturally in the classroom. The ®ndings of this study also shed some light on our initial concerns about potential negative eects of length and complexity of the content-rich DCT on intermediate level learners. Although the material did not prove to be too dicult for learners to understand, response time to Version 2 prompts was
544 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
20 minutes longer than to Version 1 prompts. For the most part Version 2 situations did not incite rampant plagiarism on the part of the learners; they did not recycle exact linguistic material from the prompt but used their own interlanguage constructions. We also noted some unanticipated bene®ts of Version 2 prompts. At least in one situation responses showed that the elaborated Version 2 prompt was clearer than its Version 1 counterpart. Although it was not the focus of our study, the issue of what constitutes the norms and appropriate pragmatic behaviour for dierent cultures is extremely important. Our ®ndings do not address the question of validity for crosscultural comparisons of pragmatic behavior, a concern raised by researchers such as Rose (1994), Rose and Ono (1995), Takahashi (1996), and Hinkel (1997). Since the non-native English speakers in this study were from seven dierent ®rst language backgrounds and cultures, it was not possible to look for such eects. Also it is not known what the eect of a `second language speaker culture'Ðthat is the culture developed by learners as they interact with each otherÐmight be on their responses. Nor did our study address the issue of how facework is perceived and carried out by the NNSs in this study. The view of facework set forth by Brown and Levinson (1987) has been challenged by recent researchers such as Scollon and Scollon (1991), and Young (1994) who maintain that although positive and negative face exist, they may be constituted quite dierently in other cultures, especially nonWestern ones. Therefore, the dierent groups of NNSs in this particular study may have been modifying their requests with goals that diered from each other as well as from those of the NSs. These are clearly limitations of this study that should be addressed in future work which looks at one ®rst language group. We do acknowledge recent debates on what the terms native speaker and non-native speaker imply, and the possible limitations of this terminology (Firth and Wagner 1997, Long 1997). We also want to make it clear that the goal of this study was not to characterize a learner language. The focus of our study was the instrument itself and whether changes to the internal structure of it would dierentially aect two populations of speakers. In conclusion, we think it is possible to strengthen design factors for which the DCT has been criticized in the past. Since it is likely that DCTs and other production questionnaires will continue to be used in sociolinguistic research, it is important to improve the quality of those studies with more carefully constructed data-gathering instruments. Until more reliable instruments are available, taking advantage of a multi-instrument approach like that developed by Hudson et al. (1995) which combines DCT, oral production, and self-assessment instruments, uses the strengths of each instrument to help oset each one's limitations. Improvements noted in this study notwithstanding, an enriched-prompt DCT has a number of limitations. It is not a good candidate for gathering data on beginning level learners unless prompts are translated into the respondent's native language. As Kasper and Schmidt (1996) point out, this weakness is of great concern for research in the development of interlanguage pragmatics. Enhanced material does take more
KRISTINE BILLMYER AND MANKA VARGHESE
545
time to respond to and respondents may suer fatigue if the questionnaire includes too many situations. Finally, even if more elaborate written prompts are better able than minimal prompts to represent some of the dimensions present in live situations, it is clear that no written instrument gathering data in a controlled setting which aords respondents the bene®t of time to plan can ever approximate the complexity, ambiguity, and ever-present unpredictability of live face-to-face interactions. Future studies will surely need to validate the major ®ndings of this study with live interactional data. Final version received February 2000
APPENDIX A Version #1 INSTRUCTIONS: Please read each situation and imagine yourself in it. Then write down what you think that you or the person in the situation would say. Use as much or as little space as you need. Finally, please answer the questions that follow each situation. 1) You are trying to study in your room and you hear loud music coming from another student's room down the hall. You don't know the student, but you decide to ask them to turn the music down. What would you say? YOU:
How imposed upon do you think the student will feel? not imposed upon moderately imposed upon 1 2 3 4 How likely do you think the student is to comply? not likely moderately likely 1 2 3 4
very imposed upon 5 very likely 5
546 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
APPENDIX B Checklist for variables Situations Content 1 2 Gender of interlocutor Social distance Role relationship Requestive goal Length of acquaintanceship Imposition/privacy (hearer's perception) Frequency of interaction (explicit) Optionality of relationship (explicit) Compliance likelihood of interlocutor Setting and scene (time, place, circumstances and psychological) Familiarity of situation Format Opportunities to respond Space Option of making request
1
2
3
4
5
6
3
4
5
6
KRISTINE BILLMYER AND MANKA VARGHESE
547
APPENDIX C Version #2 INSTRUCTIONS: Please read each situation and imagine yourself in it. Then write down what you would say. Use as much or as little space as you need. Finally, please answer the questions that follow each situation. 1) It is 10:30 p.m. on a Wednesday night and you have a paper due the next day. You are trying to ®nish the paper and you can't concentrate because you hear loud music coming from another student's room down the hall. You decide to ask her to turn the music down. The music has been on at this volume for half an hour. You have occasionally seen the student, Lucy Row, in the same dorm during the past six months. She is a student like you but you have never spoken to her. You have heard other people in the dorm complain about the volume of her music on several occasions although you never have because you usually study in the library. However today the library closed early. You are only half way through and you know that the professor for this class is very strict and does not give extensions. What would you say? YOU:
How imposed upon do you think Lucy will feel? not imposed upon moderately imposed upon 1 2 3 4 How likely do you think Lucy is to comply? not likely moderately likely 1 2 3 4
very imposed upon 5 very likely 5
548 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
APPENDIX D Situations in Version 1 1 2 3 4 5 6
You are trying to study in your room and you hear loud music coming from another student's room down the hall. You don't know the student, but you decide to ask her to turn the music down. What would you say? You missed class and need to borrow a friend's notes. What would you say? You need a ride home from school. You notice someone who lives down the street from you is also at school, but you haven't spoken to this person before. You think she might have a car. What would you say? A student in the library is making too much noise and disturbing other students. The librarian decides to ask the student to quiet down. What will the librarian say? Your term paper is due, but you haven't ®nished it yet. You want to ask the professor for an extension. What would you say? A professor wants a student to present a paper in class a week earlier than scheduled. What would the professor say?
Situations in Versions 2 1
2
3
It is 10:30 p.m. on a Wednesday night and you have a paper due the next day. You are trying to ®nish the paper and you can't concentrate because you hear loud music coming from another student's room down the hall. You decide to ask her to turn the music down. The music has been on at this volume for half an hour. You have occasionally seen the student, Lucy Row, in the same dorm during the past six months. She is a student like you but you have never spoken to her. You have heard other people in the dorm complain about the volume of her music on several occasions, although you never have because you usually study in the library. However today the library closed early. You are only half way through and you know that the professor for this class is very strict and does not give extensions. What would you say? You are at the end of a history class and you are sitting next to Tom Yates. You missed last week's class and need to borrow his notes. He has been in the same program as you for one year and you see him socially about once a month in a group. You will also be taking classes together in the future. He is a good note taker and one of the best students in the class. You have borrowed his notes twice before for the same class and the last time you borrowed them he was reluctant to give them up. In two weeks you both have the ®nal exam for your class. What would you say? It's 5:30 p.m., your last class has just ®nished and you need a ride home. You realize that a fellow classmate who was supposed to give you a ride is not in class today. You have a lot of books with you tonight, the snow has made walking dicult and you need a ride home from school. As you come out of class, you see Alice Thomas, an assistant professor in the department who teaches a class that ends at the same time as yours. She lives on the same street as you and she is standing talking to some other students. She is smiling and laughing. You have never spoken to her before but you have seen her on occasion in the department in the last few months and have both nodded to each other once or twice in the
KRISTINE BILLMYER AND MANKA VARGHESE
4
5
6
549
neighborhood. You know that she has a car and you once saw her give a lift to one of the students. What would you say? It is the end of the working day on Friday. You are a librarian and have been working in the University Reserve Room for two years. You like your job and usually the Reserve Room is quiet. Today, a student is making noise and disturbing other students. You decide to ask the student to quiet down. The student is a male student who you have often seen work on his own in the past two months, but today he is explaining something to another student in a very loud voice. A lot of students are in the library and they are studying for their midterm exams. You notice that some of the other students are looking in his direction in an annoyed manner. What would you say? Your term paper is due for a course in your major, but you haven't ®nished it yet. You want to ask the professor for an extension. You had a lot of diculty collecting data for the paper, but you think you ®nally have enough and the paper will be really good if you could have another week to put it together. Your professor is Dr. Robert Smith, senior member of the department and possibly your thesis advisor, if things go as you hope they will. You have done well in this course up to now, and he is aware of the problem with data collection. You took one course with Dr. Smith at the beginning of your studies a year and a half ago and got an A, but you haven't had much opportunity to interact with him since then. You have an appointment with Dr. Smith a few days before the paper is due. You know he rarely gives extensions on term papers because he is usually very busy, and immediately after this semester is over he will leave the campus to do ®eld work. However, you think you might have a chance because the paper is on a topic he is interested in. You are in his oce now. What would you say? You (an associate professor teaching a course in psychology) want a student to present a paper in a class a week earlier than scheduled. It is the middle of the term and topics were assigned at the beginning of the course. The presentation is a 15-minute class summary and critique of a supplementary journal article. Your student is Nancy Porter, a very competent student who always contributes to class discussions and is very well prepared for class. Even though you have never had her in class before this semester, she has a reputation as one of the best students in the department. You want her to present next week instead of three weeks from now because her article is more relevant to next week's lecture. However, midterm exams are next week and you know she has a heavy course load. She has made several contributions during this class, and has been given some good feedback from you. You ask her if you could see her for a minute after class. The students have all left and you are talking to her alone. What would you say?
550 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
APPENDIX E Request strategies Direct Clean up this mess, please. I'm asking you not to park the car here I would like you to give your lecture a week earlier. Madam, you'll have to move your car. I'd really wish you'd stop bothering me.
Mood Derivable Explicit Performative Hedged Performative Locution Derivable Scope Stating
Conventionally indirect How about cleaning up? Could you clear up the kitchen please?
Suggestory Formula Preparatory Condition
Non-conventionally indirect
You've left this kitchen in a right mess. I'm a nun. (In response to a persistent boy) (Blum-Kulka et al. 1989)
Strong Hint Mild Hint
NOTES 1 We would like to acknowledge recent work which questions the relevance of the term native speaker and assumptions of a ®xed correspondence between language and ethnicity underlying the term non-native speaker (Leung et al. 1997, Firth and Wagner 1997). We also emphasize that the goals of this study are to investigate the internal structure of the DCT as a data-collection instrument
and not to look for speci®c features of a `learner language'. NNSs here are de®ned as a group engaged in learning English in an Intensive English Program. 2 Readers should be aware that in the chi square analysis of request strategies and alerters, data do not meet the test of independence in all cases.
REFERENCES Aston, G. 1995. `Say `Thank you': Some pragmatic constraints in conversational closings.' Applied Linguistics 16/1: 57±86. Bardovi-Harlig, K. and B. Hartford. 1991. `Saying `no' in English: Native and nonnative rejections' in L. Bouton and Y. Kachru (eds.): Pragmatics and Language Learning 2: 41±58. Bardovi-Harlig, K. and B. Hartford. 1993. `Refining the DCTs; Comparing open questionnaires and dialogue completion tests' in L. Bouton and Y. Kachru (eds.): Pragmatics and Language Learning 4: 143±65. Beebe, L. M. and M. C. Cummings. 1996. `Natural speech data versus written questionnaire data: How data collection method affects speech act performance' in S. M. Gass and J. Neu (eds.): Speech Acts Across Cultures.
Berlin: Mouton de Gruyter. (Original version 1985). Beebe, L. M., T. Takahashi, and R. UlissWeltz. 1990. `Pragmatic transfer in ESL refusals' in R. Scarcella, E. Anderson, and S. Krashen (eds.): Developing Communicative Competence in a Second Language. New York: Newbury House. Bergman, M. and G. Kasper. 1993. `Perception and performance in native and nonnative apology' in G. Kasper and S. Blum-Kulka (eds.): Interlanguage Pragmatics. New York: Oxford University Press. Billmyer, K., V. Jakar, and M. J. Lee. 1989. `Developing pragmatic awareness through communicatively-oriented materials.' Paper presented at the 23rd Annual TESOL Convention, San Antonio, TX (March).
KRISTINE BILLMYER AND MANKA VARGHESE
Blum-Kulka, S. 1982. `Learning how to say what you mean in a second language: A study of the speech act performance of learners of Hebrew as a second language.' Applied Linguistics 3/1: 29±59. Blum-Kulka, S. and E. Olshtain. 1986. `Too many words: Length of utterance and pragmatic failure. Studies in Second Language Acquisition 8: 165±79. Blum-Kulka, S. and J. House. 1989. `Crosscultural and situational variation in requesting behavior' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross Cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex. 123±54. Blum-Kulka, S., J. House, and G. Kasper. 1989. Cross Cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex. Brown, P. and S. Levinson. 1987. Politeness. Cambridge: Cambridge University Press. Edmondson, W. and J. House. 1991. `Do learners talk too much? The waffle phenomenon in interlanguage pragmatics' in R. Phillipson, E. Kellerman, L. Selinker, M. Sharwood-Smith, and M. Swain (eds.): Foreign/Second Language Pedagogy Research. Clevedon: Multilingual Matters. 273±86. Eisenstein, M. and J. Bodman. 1986. ` ``I Very Appreciate'': Expressions of gratitude by native and non-native Speakers of American English.' Applied Linguistics 7/2: 157±85. Eisenstein, M. and J. Bodman. 1993. `Expressing gratitude in American English' in G. Kasper and S. Blum-Kulka (eds.): Interlanguage Pragmatics. Oxford: Oxford University Press. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford: Oxford University Press. Fasold, R. 1990. Sociolinguistics of Language. Cambridge, MA: Basil Blackwell Inc. Faerch, C. and G. Kasper. 1989. `Internal and external modification in interlanguage request realization' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross-cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex Press. 221±47. Firth, A. and J. Wagner. 1997. `On discourse, communication, and (some) fundamental concepts in SLA research.' Modern Language Journal 81/3: 285±300. Hinkel, E. 1997. `Appropriateness of advice: DCT and multiple choice data.' Applied Linguistics 18/1: 1±23.
551
Holmes, J. 1990. `Apologies in New Zealand English.' Language in Society 19/2: 155±99. Hoppe-Graff, S., T. Herrmann, P. WinterhoffSpurk, and R. Mangold. 1985. `Speech and situation: A general model for the process of speech production' in J. P. Forgas (ed.): Language and Social Situations. New York: Springer-Verlag. 81±95. House, J. 1989a. ` ``Oh excuse me please. . .'': Apologizing in a foreign language' in B. Kettemann, P. Bierbaumer, A. Fill, and A. Karpf (eds.): Englisch als Zweitsprache. Tubingen: Narr. 303±27. House, J. 1989b. `Politeness in English and German: The functions of please and bitte' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross-cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex Press. 96±119. House, J. and G. Kasper. 1987. `Interlanguage pragmatics: Requesting in a foreign language' in W. Lorscher and R. Schulze (eds.): Perspectives on Language in Performance: Festschrift for Werner Hullen. Tubingen: Narr. 1250±88. Hudson, T., E. Detmer, and J. D. Brown. 1995. `Developing prototypic measures of cross-cultural pragmatics.' (Tech. Rep. No. 7) Honolulu: University of Hawaii at Manoa, Second Language Teaching and Curriculum Center. Hymes, D. 1972. `Models of the interactions of language and social life' in J. J. Gumperz and D. Hymes (eds.): Directions in Sociolinguistics. New York: Holt, Reinhart & Winston. Johnston, B., G. Kasper, and S. Ross. 1998. `Effect of rejoinders in production questionnaires.' Applied Linguistics 19/2: 157±82. de Kadt, E. 1992. `Politeness phenomena in South African Black English.' Pragmatics and Language Learning 3: 103±16. Kasper, G. 1989. `Variation in interlanguage speech act realization' in S. Gass, C. Madden, D. Preston, and L. Selinker (eds.): Variation in Second Language Acquisition: Discourse and Pragmatics. Clevedon, UK: Multilingual Matters. Kasper, G. and M. Dahl. 1991. `Research methods in interlanguage pragmatics.' Studies in Second Language Acquisition 13/2: 215± 47. Kasper, G. and R. Schmidt. 1996. `Developmental issues in interlanguage pragmatics.' Studies in Second Language Acquisition 18/2: 149±69.
552 INVESTIGATING INSTRUMENT-BASED PRAGMATIC VARIABILITY
Labov, W. 1966. The Social Stratification of English in New York City. Washington DC: Center for Applied Linguistics. Labov, W. 1970. `The study of language in its social context.' Studium Generale 23: 30±87. Leung, C., R. Harris, and B. Rampton. 1997. `The idealised native speaker, reified ethnicities, and classroom realities.' TESOL Quarterly 31/3: 543±60. Linnell, J., F. Porter, and H. Stone 1992. `Can you apologize me? An investigation of speech act performance among non-native speakers of English.' Working Papers in Educational Linguistics 8/2: 33±53. Long, M. 1997. `Construct validity in SLA research: A response to Firth and Wagner.' The Modern Language Journal 81/3: 318±23. Manes, J. and N. Wolfson. 1981. `The compliment formula' in F. Coulmas (ed.): Conversational Routine: Explorations in Standardized Communication Situations and Prepatterned Speech. The Hague: Mouton. Meier, A. J. 1995. `Passages of politeness.' Journal of Pragmatics 24/4: 381±92. Olshtain, E. 1989. `Apologies across cultures' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross Cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex. 155±73. Olshtain, E. and A. Cohen. 1983. `Apology: A Speech Act Set' in N. Wolfson and E. Judd (eds.): Sociolinguistics and Language Acquisition. Cambridge, MA: Newbury House. Preston, D. 1989. Sociolinguistics and Second Language Acquisition. Oxford: Blackwell. Rampton, B. 1987. `Stylistic variability and not speaking ``normal'' English: Some postLabovian approaches and their implications for the study of interlanguage' in R. Ellis (ed.): Second Language Acquisition in Context. London: Prentice-Hall International. Rintell, E. and C. Mitchell. 1989. `Studying requests and apologies: An inquiry into method' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross Cultural Pragmatics:
Requests and Apologies. Norwood, NJ: Ablex. 248±72. Rose, K. 1992. `Speech act research and written questionnaires: The effect of hearer response.' Journal of Pragmatics 17/3: 49±62. Rose, K. 1994. `On validity of discourse completion tests in non-Western contexts.' Applied Linguistics 15/1: 1±14. Rose, K. and R. Ono. 1995. `Eliciting speech act data in Japanese: the effect of questionnaire type.' Language Learning 45/2: 191± 223. Scollon, R. and S. Scollon. 1991. `Topic confusion in English±Asian discourse.' World Englishes 9/2: 113±23. Takahashi, S. 1996. `Pragmatic transferability.' Studies in Second Language Acquisition 18/2: 189±223. Tarone, E. 1983. `On the variability of interlanguage systems.' Applied Linguistics 4: 143± 63. Tarone, E. 1985. `Variability in interlanguage use: a study of style-shifting in morphology and syntax.' Language Learning 35: 373±403. Wolfson, N. 1989. `The social dynamics of native and non-native variation in complimenting behavior' in M. Eisenstein (ed.): Variation in Second Language Acquisition: Empirical Views. New York: Plenum Press. 219±336. Wolfson, N., T. Marmor, and S. Jones. 1989. `Problems in the comparison of speech acts across cultures' in S. Blum-Kulka, J. House, and G. Kasper (eds.): Cross Cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex. 174±96. Young, L. 1994. Crosstalk and Culture in SinoAmerican Communication. Cambridge: Cambridge University Press. Young, R. 1988. `Variation and the interlanguage hypothesis.' Studies in Second Language Acquisition 10: 281±302. Young, R. 1991. Variation in Interlanguage Morphology. New York: Peter Lang.
Applied Linguistics 21/4: 553±571
# Oxford University Press 2000
Metaphor Awareness and Vocabulary Retention FRANK BOERS Free University of Brussels and University of Antwerp Various ®gurative expressions can often be traced back to a common metaphoric theme or source domain. The present article reports three EFL experiments which indicate that a lexical organization along such metaphoric themes or source domains can facilitate retention of unfamiliar ®gurative expressions. In view of these ®ndings, the article proposes classroom activities aimed at enhancing language learners' metaphor awareness and at turning this into an additional channel for vocabulary acquisition.
INTRODUCTION Since the publication of Lako and Johnson's Metaphors We Live By (1980) and their subsequent books outlining the paradigm of Cognitive Semantics (Johnson 1987, Lako 1987), the prevalence of ®gurative language in everyday conventional discourse has gained wide recognition. If metaphor is so omnipresent in everyday language, then language learners are bound to be confronted with ®gurative discourse at various stages of the learning process. Therefore, mastering conventional ®gurative language must be an inherent part of the language learning process too, especially since metaphors vary across cultures (KoÈvecses 1995). Moreover, many polysemous lexical items occur more frequently in their derived ®gurative senses than in their original literal senses (Low 1988). In economic discourse, for example, words like prescription and remedy are not likely to refer to real medicine. Nevertheless, the ®gurative usage of such expressions is derived from their literal senses and exploits their imagery. Drawing learners' attention to those literal senses can enhance in-depth comprehension (Boers 2000). Fortunately for the language learner, a lot of ®gurative language can be `motivated'. A wide range of ®gurative expressions can often systematically be traced back to a limited number of source domains or metaphoric themes. These recurring metaphoric themes can be employed as an alternative type of lexical ®eld, revealing structure and organization in a world of ®gurative language which may at ®rst sight seem to be largely arbitrary (KoÈvecses and Szabo 1996, Lazar 1996). The present article reports three language learning experiments that were set up to measure the potential bene®ts of organizing ®gurative expressions according to their underlying metaphoric themes. In view of their results I propose a few classroom activities aimed at raising learners' metaphor awareness and at turning this into a channel for
554 METAPHOR AWARENESS AND VOCABULARY RETENTION
vocabulary acquisition. It should be clear from the start, however, that this approach is not meant as a substitute for any established vocabulary learning methods (cf. Carter and McCarthy 1988, Hatch and Brown 1995, Huckin et al. 1993, McCarthy 1990, Schmitt and McCarthy 1997), but rather as a complementary technique. It does not constitute a separate programme, but should be conceived as being integrated with various other approaches to language teaching and learning. As an awareness-raising technique it may ®t in the broader pedagogical movement of Language Awareness (Donmall 1985, James and Garrett 1991), where language learners are encouraged not only to perform in a language, but also to re¯ect upon its use and characteristics. Since the great bulk of ®gurative language comes in multi-word expressions (idioms, proverbs, collocations, etc.), the language focus suggested here also answers recent calls to abandon the grammar/vocabulary dichotomy (Lennon 1998), to give due attention to chunk-based language (Lewis 1993), and to recognize the importance of formulaic language learning (Skehan 1998: 29±41).
THE EXPERIMENTS The three experiments described in the following sections were set up with the participation of intermediate learners of English in Belgium. Their ®rst language was either Dutch or French, i.e. languages that are rather closely related to English. While each of the experiments had a distinct language focus, the results consistently corroborate the hypothesis that a lexical organization along metaphoric themes or source domains can facilitate retention of unfamiliar ®gurative expressions. At the same time, however, the experiments point up the inevitable limitations of the approach.
Experiment 1 Participants in the ®rst experiment were 118 pupils at a Flemish secondary school, aged 16±17. Their ®rst language was Dutch, and their level of English was intermediate. Two parallel groups of pupils (taking the same course, with the same teacher) were asked to read a text, Managing the Emotions, reproduced here as Sample Text 1. Sample Text 1: Managing the Emotions People manage their emotions in dierent ways, depending on their personality and the culture they live in. In western culture, for instance, crying is usually seen as a sign of weakness, especially for men. Research has shown, however, that crying it out makes people feel better. So perhaps we should encourage our children to wear their hearts on their sleeves more often. It has become a widely accepted idea that, instead of bottling up the emotions, one should ventilate one's emotions once in a while. One should blow o steam in order to avoid sudden explosions. However, one emotion may prove to be an
FRANK BOERS
555
exception: anger. In fact, ventilating anger is one of the worst ways to cool down: outbursts of rage typically pump up the arousal, leaving people feeling more angry, not less. When people blow up at the person who has provoked their anger, their rage not only peaks during the outburst, but the angry mood is also prolonged. After reaching their boiling point, these people keep fuming much longer. A far more eective way of managing anger is to simmer down ®rst, and then, in a more constructive manner, confront the person to settle a dispute. Imagine that someone pushes you aside in a crowd. Your ®rst thought may be `How rude !' That re¯ex may then be followed by more negative thoughts: `He could have hurt me! I can't let him get away with that!' Then, should someone else behind you bump into you because you have slowed down, you are apt to erupt in rage at that person too. Now imagine a more positive line of thought toward the man who pushed you aside: `Maybe he had a good reason, such as an emergency'.
This text (contrived for teaching purposes) was inspired by Goleman's Emotional Intelligence: why it can matter more than IQ (1995: 62±72). It exempli®es the metaphoric themes THE BODY IS A CONTAINER FOR EMOTIONS and, more speci®cally, ANGER IS A HOT FLUID IN A CONTAINER. After reading the text, the experimental group (24 girls and 34 boys) received vocabulary notes organized along various metaphoric themes (as identi®ed by KoÈvecses 1986, 1990), reproduced here under Sample Lexis 1a. The control group (28 girls and 32 boys) received the same vocabulary input organized along dierent (pragmatic or functional) lines, reproduced here under Sample Lexis 1b. Sample Lexis 1a: English has a lot of expressions to describe anger. Some of these are very common: She's angry. He's mad at you. To make your language more varied, other expressions can be used to specify the kind of anger: anger as a hot ¯uid in a container anger welled up inside me simmer down! I was boiling with anger she ¯ipped her lid she was all steamed up I was fuming she erupted she blew up at me anger as ®re an in¯ammatory remark she was breathing ®re adding fuel to the ®re she exploded he kept smouldering for days he's hot under the collar angry people as dangerous animals he has a ferocious temper he unleashed his anger don't snap at me! don't bite my head o! Sample Lexis 1b: English has a lot of expressions to describe anger. Some of these are very common: She's angry. He's mad at you. To make your language more varied, other expressions can be used to specify the kind of anger:
556 METAPHOR AWARENESS AND VOCABULARY RETENTION
to describe acute and sudden anger she exploded he unleashed his anger to describe anger as a process anger welled up inside me I was boiling with anger He was fuming I kept smouldering for days to describe angry personalities he's hot under the collar to describe the way angry people speak she blew up at me don't bite my head o!
she erupted she ¯ipped her lid simmer down! she was all steamed up an in¯ammatory remark adding fuel to the ®re he has a ferocious temper don't snap at me! she was breathing ®re
The participants were given 10 minutes to look over the vocabulary and to ask for any further clari®cation. Subsequently they engaged in a guided class discussion (15 minutes) about anger and con¯icts. The vocabulary notes were removed and ®nally the participants were given a cloze test (reproduced here as Sample Text 2), with ten items meant to elicit the lexis studied. Sample Text 2: Fill in the gaps with one word each. Sometimes there may be more than one option. In that case, try to come up with as many possibilities as you can. Last month was my parents' wedding anniversary. A week before the anniversary my mother already suspected that my father had forgotten about it, since he hadn't asked her if she wanted to do anything special for the occasion. After all those years my father should really have known better, because he had learned from (1) temper my mother has. But I felt experience what a that he deserved another lesson and I decided not to remind him of the anniversary. Days before that fatal day I could already notice my (2) up inside her. I tried to tell her mother's anger (3) down, because my father might still remember after to all. But he didn't, and as the day of the anniversary approached, (4) up. On the morning of the day mother was getting all itself, my father left for work as usual without mentioning the (5) her lid. She had wedding anniversary. Mother almost (6) point and she kept fuming all day. clearly reached her To make matters even worse, father was late that evening. By the time he got home, mother was ready to explode. And when he ®nally got home at 8 pm without as much as a present for her, (7) her anger. `You inconsiderate, sel®sh oaf!' she (8) at him. My father looked baed. `I know I'm a she bit late, but that's no reason to bite my head o,' he mumbled. I felt this was the perfect time to make my contribution to their marital bliss. I took the expensive bottle of champagne that I had bought and oered it to them, congratulating them on their wedding anniversary. (9) to the ®re. My For my mother this, of course, added
FRANK BOERS
557
father, who looked absolutely embarrassed now, made a desperate attempt to make up by taking her out for dinner at her favourite restaurant, but even so she kept breathing ®re all night. He ended up buying several uninspired presents and a bouquet of roses the next (10) for weeks. day to make her cool down, but she kept
The pupils were encouraged to list several possibilities per gap if they could. Appropriate responses for the third item, for example, included calm down and cool down as well as the targeted expression simmer down. Since the aim of the experiment was to measure the eect of metaphor awareness on learners' retention of novel vocabulary, the answer sheets were screened primarily for the reproduction of the lexis that the pupils had previously been asked to look over.
Results The participants who had received the vocabulary notes organized along metaphoric themes were more likely than the control group to reproduce the lexis studied (p < .05). On average they responded to 4.41 of the 10 gaps with the targeted words. The average score of the control group was 3.67. The results of this ®rst experiment suggest that, for the domain of emotions at least, an awareness of the metaphoric themes behind novel vocabulary can facilitate retention. We should bear in mind, however, that the metaphoric themes along which the vocabulary input was organized in the experiment also existed in the participants' L1. This may have helped the experimental group's learning process, because transfer from L1 to the target language can speed up the learning process (at least when the two languages share many features). However, a transfer strategy inevitably involves the risk of erroneous L1 interference as well (Swan 1997). According to Kellerman (1987), the use of transfer strategies is most likely when the language learner perceives the two languages to be `close'. While this may facilitate the acquisition of idioms through association with a metaphoric theme that is shared by the two languages, it also raises the risk of erroneous `direct' translations (Cornell 1999). After all, the instantiations of shared metaphoric themes vary across languages. The Dutch equivalent of Biting someone's head o, for example, is `Biting someone's nose o'. The ninth item of the cloze test was meant to elicit the expression To add fuel to the ®re. The equivalent expression in Dutch is `To add oil to the ®re'. Probably due to L1 interference, nine pupils in the experimental group and three pupils in the control group responded with oil.
Experiment 2 A cloze test like the one used in the ®rst experiment is by de®nition pretty much controlled and the respondent's role is fairly passive. Therefore a second experiment was set up to measure the eect of metaphor awareness on the
558 METAPHOR AWARENESS AND VOCABULARY RETENTION
reproduction of novel vocabulary in more active usage. Participants in this experiment were 73 university students of business and economics, aged 19± 20. Their ®rst language was French and their level of English was intermediate. As part of their common English course, the participants were given a list of vocabulary (reproduced here under Sample Lexis 2) to enable them to describe upward and downward economic trends. Sample Lexis 2: UPWARD TRENDS DOWNWARD TRENDS (t = transitive; i = intransitive ) Verbs Nouns Verbs Nouns increase (t/i) increase decrease (t/i) decrease rise (i) rise fall (i) fall grow (i) growth shrink (i) raise (t) drop (i) drop put up (t) put down (t) push up (t) push down (t) soar (i) decline (i) decline surge (i) surge cut (t) cut plunge (i) plunge peak (i) peak dive (i) dive perk up (i) go downhill (i) mount (i) drive down (t) creep up (i) plummet (i) slide (i) slide crash (i) crash
All participants also received the following introductory lines with their vocabulary list: Instead of resorting to simple and rather vague expressions like go up and go down, you can make your descriptions of upward and downward trends more varied by using verbs and expressions that oer a more precise picture.
For the experimental group (13 female and 27 male students) the introduction continued as follows: Some expressions call up a speci®c image, such as rockets or airplanes: `soar', `skyrocket', `crash'; diving: `plunge', `dive'; mountain climbing: `mount', `creep up', `go downhill', `slide', `peak'.
By drawing the students' attention to the source domains of the given expressions, it was hoped that they would be encouraged to apply imagery in their processing of the word list. For the control group (15 female and 18 male students), on the other hand, the second part of the introductory note read as follows: Some expressions indicate the speed of change, such as fast change: `soar', `skyrocket', `plunge', `dive';
FRANK BOERS
559
gradual change: `creep up', `mount', `slide', `go downhill'; or reaching a limit: `peak', `crash'.
The participants were given 10 minutes to go over the word list and to ask for any clari®cation. Then they were shown a couple of graphs depicting various countries' economic growth and unemployment ®gures. The vocabulary notes were removed and the participants were given about 30 minutes to write a short essay describing the graphs on display. The task was introduced to the students as a revision exercise on the tenses (the graphs represented past, present, and future trends), but the students were also invited to vary their up±down lexis. Afterwards the written descriptions were collected and scanned for the items included in the wordlist. Verb-noun equivalents (e.g. a diveÐto dive) were counted as a single occurrence. Incorrect uses of the given expressions were excluded. Such inaccuracies included mixing up transitive and intransitive verbs (e.g. rise instead of raise), wrong morphology (e.g. skyrocked instead of skyrocketed), and semantic incoherence (e.g. gradual plunge; slightly soaring; sliding from 5% to 10%).
Results The average number of targeted expressions used by the experimental group was 7.1 compared to 4.9 for the control group. The participants who had been encouraged to process the ®gurative items in association with their source domains were much more likely (p < .001) to reproduce them in active usage. This ®nding holds for the lexical items that were merely listed in the table (without explicit imagery) as well as those highlighted in the introductory paragraph. With respect to the overall number of inaccurate uses of the targeted expressions, both groups were on a par. Still, it may be interesting to note that ®ve essays produced by the control group contained a case of semantic incoherence, compared to only one in the experimental group.
Experiment 3 The expressions that were the language focus of the second experiment all instantiate the general orientational metaphor MORE IS UP; LESS IS DOWN. Orientational metaphors underlie many prepositional and phrasal verbs (multi-word verbs), which are often considered to be a major obstacle on the road to English pro®ciency. In this journal, KoÈvecses and Szabo (1996) reported an EFL experiment about the possible merits of a cognitive semantic approach to teaching and learning phrasal verbs, i.e. an approach which raises learners' awareness of the conceptual metaphors behind ®gurative language. The results of their study were promising, but unfortunately the scale of the experiment was too limited for statistical analysis. In order to collect more conclusive data I set up a similar experiment with the collaboration of a greater number of participants.
560 METAPHOR AWARENESS AND VOCABULARY RETENTION
This third experiment involved 74 university students, aged 19±20. Their ®rst language was French and their level of English was intermediate. As part of their common English course, the students were presented with a set of prepositional and phrasal verbs, selected from the list provided in A Practical English Grammar (Thomson and Martinet 1980: 295±339). The control group (13 female and 22 male students) received explanatory notes on the multiword verbs as presented in A Practical English Grammar, i.e. listed alphabetically. The experimental group (19 female and 20 male students) received the same input, with the exception that here the multi-word verbs were categorized under the headings of their underlying orientational metaphors. This presentation (reproduced here as Sample Lexis 3) was based on lexicosemantic analyses of prepositions and phrasal verbs by Boers (1996), Lindner (1981) and Lindstromberg (1997). Sample Lexis 3 MORE IS UP; LESS IS DOWN * blow up = in¯ate, pump up, exaggerate * cut down (prices, expenses, taxes, etc.) * turn up/down (the radio, the central heating, etc.) ACTIVE IS UP; INACTIVE IS DOWN * set up (a business, an experiment, etc.) = create * break down = collapse, stop functioning ? emotional breakdown * close down/shut down (a factory or business) GOOD IS UP; BAD IS DOWN * be down/feel down = be unhappy cheer up = become happy * feel up to a certain task = feel strong enough VISIBLE IS OUT and UP; INVISIBLE IS IN and DOWN * come up with an idea, a solution = propose, mention. * ®nd out something = discover. * ®gure out/work out a problem = solve a problem * look up something in a dictionary, etc. * point out something = indicate, show * show up/turn up = arrive, appear * turn out products = produce turn out = develop, appear, conclude (e.g. The initiative turned out to be a big success) IMAGERY OF MANIPULATING OBJECTS * give up something = abandon an attempt, stop a habit. * put forward a proposal = suggest. * take over = take control or responsibility. * take up (a hobby, profession or study) = start it.
Both groups of participants were given 10 minutes to study the set of multiword verbs and to ask for any further clari®cation. Subsequently the vocabulary notes were removed and the participants were asked to do a
FRANK BOERS
561
cloze test. This gap®ll exercise (reproduced here under Sample Text 3) was inspired by Goleman's Emotional Intelligence (1995: 91±4) again. Sample Text 3 The marshmallow test Just imagine you're four years old, and an adult makes the following proposal: If you wait until after he runs an errand, you can have two marshmallows for a treat. If you can't wait until then, you can only have oneÐbut you can have it right now. This is a dilemma: to to impulsive desire or to delay grati®cation. This (1) by psychologists in the remarkable experiment was (2) 1960s. Some of the four-year-olds were able to wait what must surely have seemed an endless ®fteen to twenty minutes for the experimenter to return. These children got the two-marshmallow reward. But others, more impulsive, grabbed the one marshmallow, almost always within seconds of the experimenter's leaving the room. The diagnostic power of the test became clear some twelve to fourteen years later, when as adolescents. The emotional these same children were (3) and social dierence between the grab-the-marshmallow children and to be dramatic. The the grati®cation-delaying ones (4) that those who had resisted temptation researchers (5) frustrations. They were, as adolescents, better able to (6) under stress. They (8) were less likely to (7) challenging activities and pursued them instead of (9) in the face of diculties. And, more than a decade later, they were still able to delay grati®cation in pursuit of their goals. Those who almost immediately grabbed for the marshmallow, however, tended to have fewer of these qualities. In adolescence they found it harder to their minds about dicult choices. They were more (10) easily upset by frustrations. They lacked con®dence and did not challenging tasks. They were more likely to be jealous (11) . and consequently, they were typically dicult to (12) And, after all those years, they were still unable to (13) grati®cation. What (14) in a small way early in life . The blossoms into a wide range of abilities as life (15) capacity to impose a delay on impulse is at the root of a plethora of eorts, from staying on a diet to pursuing a university degree. Some children had already mastered the basics at the age of four: they that in the given situation delay was bene®cial and (16) they had the necessary perseverance toward their goalÐthe two marshmallows. Even more surprising, when the tested children were evaluated again as they were ®nishing high school, those who had waited patiently at four were far superior as students to those who had acted on whim. According to their parents' evaluations, they were their more academically competent and better able to (17) plans and studies. Those who had given in to impulse at four were school and to (19) . The more likely to (18) that the marshmallow test is twice results of the study (20) as powerful a predictor of academic achievements as an IQ test.
562 METAPHOR AWARENESS AND VOCABULARY RETENTION
The participants were given 20 minutes to choose from the following list of multi-word verbs to complete 20 gaps in the text: put o; cope with; giving up; drop out; shows up; feel up to; took up; goes on; be fed up with; set up; break down; make up; ®gured out; get on with; give in; point out; follow through with; turned out; tracked down; found out.
The appropriate morphology of the verbs was provided to facilitate the task. Half of the multi-word verbs to be chosen from were included in the previously studied vocabulary notes, while the other half were not. The latter were incorporated into the exercise to investigate the possibility of successful transfer of the strategy of spatial imagery (see below).
Results Let us ®rst consider the results pertaining to the ten items explained in the participants' vocabulary notes. The average score of the experimental group on these was 5.65, compared to 4.23 for the control group. The participants who had studied the multi-word verbs categorized under orientational metaphors proved more likely (p < .01) to correctly ®ll in the gaps meant to elicit these items. This result con®rms the basic trend noted by KoÈvecses and Szabo (1996: 349±50). The results of their experiment also suggested successful transfer of the cognitive semantic approach when language learners try to tackle novel phrasal verbs (KoÈvecses and Szabo 1996: 351). The present experiment, however, oers no support to that suggestion. The experimental group did not perform any better than the control group on the ten gaps meant to elicit the multi-word verbs that had not been included in their vocabulary notes. The average scores on these were 4.07 and 4.2, respectively. This result shows that the experimental group did not bene®t much from their enhanced awareness of certain orientational metaphors in their dealings with novel multi-word verbs. These may instantiate metaphors (or clusters of metaphors) that were absent from the initial vocabulary list (e.g. PATH metaphors in go on and follow through with). Furthermore, phrasal and prepositional verbs vary in their degrees of semantic transparency. While some are fairly easily imageable and guessable (Lindstromberg 1997: 17±20), many others may turn out to be too opaque to lend themselves to straightforward imagery processing.
DISCUSSION The three learning experiments reported above suggest that language learners' lexical resources bene®t from an enhanced metaphor awareness. In each of the experiments the experimental group, which had been presented with ®gurative expressions organized along their underlying metaphoric themes, revealed superior retention.
FRANK BOERS
563
Several conditions may jointly have contributed to these superior results (cf. Ellis 1994; SoÈkmen 1997): 1 2 3
Learning vocabulary through imagery processing (in addition to verbal processing) paves an extra pathway for later recall. Employing cognitive eort to identify source domains and to make categorization judgements promotes deep-level cognitive processing, which in turn promotes memory storage. Applying metaphoric themes as categories provides a framework for lexical organization, and organized vocabulary is known to be easier to learn than random lists.
The generally encouraging ®ndings still need to be interpreted within the con®nes of the study, of course. First, most of the ®gurative expressions focused on in the experiments were semantically rather transparent or imageable. Metaphor awareness will most probably be less fruitful when the learner is faced with opaque idioms. The degree of semantic transparency of a ®gurative expression is determined by the interplay of various factors (cf. Flores d'Arcais 1993, Gibbs 1993). One of those factors seems to be its association with an established metaphoric theme. Expressions which re¯ect such a metaphoric theme tend to be more transparent than more `isolated' cases. For example, To let o steam (which, along with She was fuming, He got all steamed up, She erupted, etc., re¯ects the metaphoric theme ANGER IS HEAT) may be more transparent than To sell someone down the river (which does not seem to belong to a cluster of expressions re¯ecting a metaphoric theme). Furthermore, expressions that are closely associated with a given metaphoric theme tend to be more transparent than more `peripheral' ones. For example, She was fuming is a more `central' instance of ANGER IS HEAT than She ¯ipped her lid or He hit the ceiling (Lako 1987: 384±5). Secondly, the ®rst language of the participants (Dutch or French) in the three experiments was rather closely related to the target language (English). The metaphoric themes behind the given ®gurative expressions were also common in the learners' L1. This may have facilitated comprehension. Learners of a `distant' language, however, may face comprehension problems due to cross-cultural dierences in conventionalized metaphoric themes. For example, She broke my heart may be semantically quite opaque to members of a community whose culture does not conceive of the heart as the seat of the emotions (cf. Chitra 1996: 124±35). On the other hand, a perceived `closeness' of the L1 and the target language may prompt learners to (over-)use transfer strategies, which may then result in erroneous direct translations. Thirdly, the results of the experiments pertain to participants whose level of English was intermediate. Beginners' comprehension of many ®gurative expressions would clearly be impeded by a lack of lexical resources. For example, in order to recognize She was fuming and He hit the ceiling as ®gurative expressions that re¯ect ANGER IS HEAT, one obviously needs to comprehend
564 METAPHOR AWARENESS AND VOCABULARY RETENTION
the words fuming and ceiling. Advanced learners, on the other hand, are less likely to face such obstacles and would hence be in a position to bene®t most from their enhanced awareness of metaphor. Moreover, as advanced learners tend to be hesitant about the transferability of L1 idioms to the target language (Kellerman 1978), they may also be more cautious about direct translations. Although two linguistic cultures may share an established metaphoric theme, the way this theme is instantiated through conventional linguistic expressions may vary considerably. It may be feasible to relate a set of idiomatic expressions to a common metaphoric theme or source domain, but it remains impossible to predict exactly what the idioms belonging to that metaphoric theme will look like in a particular language. As a result, metaphor awareness is not meant to be used by the learner as a `generator' of the conventional ®gurative expressions of the target language. Instead, its primary use proposed here is as a channelling device for learners to organize the steady stream of ®gurative language they are exposed to. Finally, it should be mentioned that the three reported experiments provide no information as to what psychological pro®les bene®t most from raising metaphor awareness. Individual learners may respond to the treatment in dierent ways, depending on their particular cognitive styles. Neither were the experiments controlled for any potential gender dierences in metaphor processing. In the three EFL experiments, the participants were presented with prepackaged lexical input, already categorized under identi®ed metaphoric themes. Metaphor awareness can only be fruitful in the long term, however, if learners are capable of identifying metaphoric themes and of categorizing idioms independently. To test this ability in learners I carried out the following small-scale experiment. The participants were 64 university students, once again of business and economics (27 female and 37 male students, aged 19±20, with French as L1, and with an intermediate level of English). To my knowledge the participants had not received any metaphor instruction before and, in addition, on this occasion it was kept minimal and simple. The participants were given a questionnaire with the following introduction: Economic processes are described by means of a variety of metaphoric expressions: taris are `trade barriers', money transfers are `cash¯ow', employees are `human capital', starting ®rms are `infant companies', stockmarkets may `crash', demand for a product may be `elastic', and so on. Quite often, various ®gurative expressions relate to a single source of inspiration: `The exchange rate mechanism' and `Using the right monetary tools' are both inspired by MACHINERY. `Healthy ®rms' and `Sickly companies' are both inspired by HEALTH. `A takeover battle' and `Protectionist measures' are both inspired by WAR. And so on.
Subsequently, the participants were given 5 minutes to categorize ®fteen expressions under MACHINERY, HEALTH, WAR, or a fourth category which
FRANK BOERS
565
they would have to identify themselves. The expressions (which were listed in random order) were: 1 2
3 4
as instances of the MACHINERY metaphor: To tighten the screws on the economy; The economy is overheating; To ®ne-tune in¯ation; The monetary lever has rusted; as instances of the HEALTH metaphor: A chronic budget de®cit; Symptoms of an arthritic labour market; To prescribe the right economic remedy; The company will have to slim down; The economy is slowly recovering; as instances of the WAR metaphor: To invade weaker markets; The right strategy to penetrate the Russian market; To conquer market share; the fourth category, which could come under the heading of GARDENING, was represented by: To get to the roots of a thorny problem; A ¯ourishing company; The company will prune some of its branches.
The expressions and the proposed categorization were borrowed from analyses of economic discourse by Henderson (1986) and Boers and Demecheleer (1997). The participants were asked to underline any words they did not understand. After all, they could not be expected to recognize the metaphoric theme behind an expression (e.g. MACHINERY), if they did not comprehend the key words that constituted the expression (e.g. Tighten the screws). Overall, 16.5 per cent of the responses were blocked by a lack of such lexical knowledge. This left us with a total of 801 categorization judgements.
Results In all, 89.5 per cent of the respondents' categorization judgements corresponded to our own analysis. This suggests that learners are indeed capable of grouping suciently transparent ®gurative expressions under given metaphoric themes. The three expressions included in the questionnaire as instances of the GARDENING metaphor were meant to measure the learners' ability to identify a metaphoric theme independently. Again, the results were encouraging: 75.4 per cent of the responses on these three items proposed GARDENING, VEGETATION or NATURE as the `source of inspiration'. Several judgements that diered from our own categorization re¯ected the respondents' recognition of additional `potential' source domains. The economy is overheating, for example, was associated by some participants with COOKING. To ®ne-tune in¯ation was associated by some respondents with MUSIC. The right strategy to penetrate the Russian market was associated by some with SEX. Some confusion may have occurred due to L1 interference. A few respondents appeared to have mistaken the verb prune in The company has to prune some of its branches for the equivalent of the French noun prune (which means `plum') and hence assumed the `source of inspiration' was FOOD. In
566 METAPHOR AWARENESS AND VOCABULARY RETENTION
addition, twelve participants may have been led to categorize A ¯ourishing company under the HEALTH metaphor because of the French idiom Une sante ¯orissante (`a ¯ourishing health'). Identifying metaphoric themes in actual discourse and categorizing ®gurative expressions along those themes raises many methodological problems and is clearly not a matter of clear-cut choices (Boers 1997, 1999). Dierent metaphoric themes intersect, so that dierent ones may be at play simultaneously (e.g. HEALTH, FITNESS and SPORTS metaphors). In addition, rather speci®c metaphoric themes (e.g. GARDENING) can be subsumed under more generic ones (e.g. NATURE). Consequently, a fair degree of variation in language users' interpretation of ®gurative language is to be expected (and perhaps even to be encouraged for pollination and awareness-raising purposes).
ENHANCING METAPHOR AWARENESS IN THE CLASSROOM The general aim of raising language learners' awareness of metaphor can be broken down into the following more speci®c objectives: (i) recognition of metaphor as a common ingredient of everyday language; (ii) recognition of metaphoric themes behind many ®gurative expressions; (iii) recognition of the non-arbitrary nature of many ®gurative expressions; (iv) recognition of possible cross-cultural dierences in metaphoric themes; and (v) recognition of cross-linguistic variety in ®gurative expressions. In the present section I shall proceed by describing a couple of awareness-raising activities and the way these have been successfully piloted in classroom practice. One way of making learners aware that metaphor is a very common ingredient of everyday language is to ask them to consider their own language about an abstract phenomenon. A sample activity that has proved its worth in classroom practice is to have students de®ne the dierence between friendship and love. As these are abstract concepts, most of the proposed de®nitions will have a metaphoric underpinning. Love and friendship will be likened to concrete source domains, each of which de®nes a metaphoric theme. The metaphoric themes that are often generated by this activity include spatial metaphors (e.g. Love is deeper than friendship, while friendship is more shallow), business metaphors (e.g. Love is an exclusive deal, while you can share many friends), architectural metaphors (e.g. Love is based on aection, while the cornerstone of friendship is trust), body-part metaphors (e.g. Love is a matter of the heart, while friendship is a matter of the mind), and so on. By drawing students' attention to the ®gurative nature of their own de®nitions, they will realize that metaphor is not just an ornamental device con®ned to poetry, but rather a typical aspect of language (and thought) in general. The `love versus friendship' exercise can easily be integrated, for example as part of a wider thematic project around gender dierences. Men and women may have dierent perspectives on emotion concepts, and such dierences may be re¯ected in their discourse. This turned out to be the case when I gave 42 female and 31 male university students 10 minutes to write down their
FRANK BOERS
567
de®nitions of love and friendship. Their short essays were then screened for ®gurative expressions re¯ecting dierent metaphoric themes (i.e. building on dierent source domains). On average, the female students used a wider variety of metaphoric themes, while the male students typically stuck to one metaphoric theme to explain their views: 40.5 per cent of the female students used at least three dierent metaphoric themes in their short essays, while only 22.5 per cent of the male students did so. For now, this quantitative gender dierence in metaphor production should be interpreted solely in connection with this particular task and the emotion concepts at hand. Although it must be said that they were set up without any control for such potential dierences, none of the other experiments reported in this article revealed any gender dierences in metaphor processing: the female and male participants were on a par in the three vocabulary retention tests as well as the categorization exercise. The `love versus friendship' essays also showed dierent preferences regarding metaphoric themes. While 26 per cent of the girls described love/ friendship in terms of `sharing' feelings and secrets, only 13 per cent of the boys did so. Conversely, 42 per cent of the male students used architectural imagery (love/friendship as a construction that has to be built or can be demolished), compared to only 19 per cent of the female population. Rather surprisingly, no fewer than 26 per cent of the girls mentioned physical `attraction' as characteristic of love, whereas only 6 per cent of the male population did so. These observations were subsequently recycled to fuel a debate around gender dierences. The exercise illustrates how focusing on the ®gurative language produced by language learners themselves may serve as a means of enhancing their metaphor awareness while still being integrated in wider thematic projects and communicative activities. A second aim of enhancing metaphor awareness is for the learner to recognize that the wide variety of ®gurative expressions she or he is confronted with need not be tackled as random lists. Instead, many idioms can be grouped under more general metaphoric themes (or their source domains), as was illustrated by the experiments. The existence of these metaphoric themes can then be `explained' through reference to their experiential basis. As a pre-reading activity to Sample Text 1 (Managing the Emotions), for instance, students could be asked to list the symptoms of anger. When angry, people typically become red in the face, they become irrational, etc. Each of these symptoms feeds into a dierent metaphoric conception of the emotion of anger (ANGER IS HEAT, ANGER IS INSANITY, etc.), as re¯ected in ®gurative language. `Motivating' the use of a given metaphoric theme by referring to its correlate in physical experience may also improve language learners' in-depth comprehension of its linguistic instances. Because of the `logic' of the imagery of heating up a ¯uid in a container, stewing or simmering express a dierent stage in the process than bursting or erupting, for example. As a result, one would not describe a mildly irritated person as being on the point of bursting. In other words, the `logic' of
568 METAPHOR AWARENESS AND VOCABULARY RETENTION
the metaphor helps the language user choose its appropriate instantiations to ®t a given context (cf. Gibbs 1993). In addition to explaining general metaphoric themes, language learners may also be asked to try to explain individual idiomatic expressions that have a sucient degree of semantic transparency. To keep something under one's hat, for instance, exploits the metaphors KNOWING IS SEEING and THE MIND IS A CONTAINER. The hat covers the skull (i.e. the top of the container) and hence hides its contents from view. This exercise may seem challenging, but it appears not to be beyond language learners' competence. When I asked 78 French-speaking students to guess the meaning of keeping something under one's hat without any contextual clues, 47 per cent of the participants gave a correct response, despite their claim that they had never encountered the expression before and despite the absence of a close equivalent in their native tongue. Learners can be encouraged to ®rst try to decode ®gurative expressions independently, i.e. as a problem-solving task requiring a deeper level of cognitive processing, before resorting to the teacher or the dictionary for corroboration or falsi®cation (Lennon 1998). Moreover, in normal learning conditions idioms are encountered in context, which facilitates comprehension considerably (Cooper 1999). The lower the degree of semantic transparency of the expression, the more the learner will have to rely on contextual clues to ®gure out its meaning. Imagery processing and employing contextual clues can nevertheless be mutually supportive strategies to guess the meaning of unfamiliar words. When a metaphoric theme is spread over a stretch of discourse as a cohesive device, the activation of the source domain through a known word may help the learner guess the meaning of an unfamiliar word in its proximity. For example, if the known word medicine calls up the source domain of HEALTH, then the learner may ®nd it easier to guess the meaning of prescribe in A good economist should prescribe the right economic medicine. Similarly, if the known word branches activates the source domain of GARDENING, then the learner might be in a better position to work out the meaning of prune in The company will have to prune several of its branches. Idioms can sometimes be taken as a re¯ection of historical-cultural backgrounds. The frequently used imagery of the hat in various English idioms (e.g. Pass the hat round, Talk through one's hat, Hang up one's hat), for example, may re¯ect part of the national stereotype of the English gentleman with his bowler hat and walking-stick. Similarly, the high frequency and variety of SHIP metaphors in English may be anchored in Britain's geography and history. This historical-cultural perspective can further be exploited by having language learners compare the ®gurative discourse of the target language with their L1 (e.g. Boers and Demecheleer 1998; Deignan, Gabrys, and Solska 1997). On a conceptual level, such an enterprise may reveal crosscultural dierences in conventionalized metaphoric themes. On a linguistic level, it may lay bare the risk of L1 interference and erroneous direct translations.
FRANK BOERS
569
CONCLUSION In this article I have reported three learning experiments which corroborate the thesis that an enhanced metaphor awareness on the part of the language learner can facilitate her or his retention of novel ®gurative expressions. In view of these ®ndings, I have then described a small number of classroom activities aimed at enhancing learners' metaphor awareness and at turning this into an additional channel for vocabulary acquisition. In this proposal, the metaphoric themes or source domains behind multiple ®gurative expressions constitute a useful framework for lexical organization. At the same time, however, we have had to acknowledge a number of limitations along two dimensions: 1 2
Not all ®gurative language lends itself equally well to the approach. Certain idioms may be too opaque and thus not imageable enough. Other idioms may be hard to capture under any identi®ed metaphoric theme. Knowledge of the conventional metaphoric themes of a given language does not guarantee mastery of its conventional linguistic instantiations. As it is impossible to predict exactly how a particular language will instantiate identi®ed metaphoric themes, learners cannot employ their awareness of those metaphoric themes to `generate' ®gurative expressions in the target languageÐunless they wish to produce original or poetic language.
Further research would be indispensable to determine whether these limits can be stretched. Final version received February 2000
ACKNOWLEDGEMENTS I would like to thank the editors and the anonymous reviewers for their helpful advice and insightful comments on an earlier version of this paper. Also many thanks to my colleagues and to several generations of students for putting up with my obscure experiments.
REFERENCES Boers, F. 1996. Spatial Prepositions and Metaphor: A Cognitive Semantic Journey along the Up±Down and the Front±Back Dimensions. TuÈbingen: Gunter Narr Verlag. Boers, F. 1997. `No pain, no gain in a free market rhetoric: a test for cognitive semantics?' Metaphor and Symbol 12/4: 231±41. Boers, F. 1999. `When a bodily source domain becomes prominent: the joy of counting metaphors in the socio-economic domain' in R. W. Gibbs and G. Steen (eds.): Metaphor in Cognitive Linguistics. Amsterdam/Philadelphia: John Benjamins. 47±56.
Boers, F. 2000. `Enhancing metaphoric awareness in specialised reading.' English For Specific Purposes, 19/2: 137±47. Boers, F. and M. Demecheleer. 1997. `A few metaphorical models in (western) economic discourse' in W. A. Liebert, G. Redeker and L. Waugh (eds.): Discourse and Perspective in Cognitive Linguistics. Amsterdam/Philadelphia: John Benjamins. 115±129. Boers, F. and M. Demecheleer. 1998. `A cognitive semantic approach to teaching prepositions.' ELT Journal, 52/3: 197±203. Carter, R. and M. McCarthy. (eds.) 1988.
570 METAPHOR AWARENESS AND VOCABULARY RETENTION
Vocabulary and Language Teaching. London: Longman. Chitra, F. 1996. Idioms and Idiomaticity. Oxford and New York: Oxford University Press. Cooper, T. 1999. `Processing of idioms by L2 learners of English'. TESOL Quarterly 33/2: 233±62. Cornell, A. 1999. `Idioms: an approach to identifying major pitfalls for learners'. IRAL 37/1: 1±21. Deignan, A., D. Gabrys, and A. Solska. 1997. `Teaching English metaphors using crosslinguistic awareness-raising activities' ELT Journal, 51/4: 352±60. Donmall, B. G. (ed.). 1985. Language Awareness: NCLE Reports and Papers 6. CILT, London. Ellis, N. C. 1994. `Vocabulary acquisition: the implicit ins and outs of explicit cognitive mediation' in N. C. Ellis (ed.): Implicit and Explicit Learning of Languages. London/San Diego: Academic Press. 211±82. Flores d'Arcais, G. B. 1993. `The comprehension and semantic interpretation of idioms' in C. Cacciari and P. Tabossi (eds.): Idioms: Processing, Structure, and Interpretation. Hillsdale/London: Lawrence Erlbaum Associates. 79±98. Gibbs, R. W. 1993. `Why idioms are not dead metaphors' in C. Cacciari and P. Tabossi (eds.): Idioms: Processing, Structure, and Interpretation. Hillsdale/London: Lawrence Erlbaum Associates. 57±76. Goleman, D. 1995. Emotional Intelligence: Why it can matter more than IQ. New York: Bantam Books. Henderson, W. 1986. `Metaphor in economics' in M. Coulthard (ed.): Talking about Text (Discourse Monograph No 13). Birmingham: English Language Research. 109±27. Hatch, E. and C. Brown. 1995. Vocabulary, Semantics, and Language Education. Cambridge: Cambridge University Press. Huckin, T., M. Haynes, and J. Coady. (eds.) 1993. Second Language Reading and Vocabulary Learning. Norwood: Ablex. James, C. and P. Garrett. (eds.) 1991. Language Awareness in the Classroom. London/New York: Longman. Johnson, M. 1987. The Body in the Mind: The Bodily Basis of Meaning, Imagination and Reason. Chicago/London: University of Chicago Press. Kellerman, E. 1978. `Giving learners a break:
native language intuitions as a source of predictions about transferability.' Working Papers in Bilingualism 15: 59±92. Kellerman, E. 1987. `Aspects of transferability in second language acquisition'. Unpublished Ph.D. Dissertation, University of Nijmegen. KoÈvecses, Z. 1986. Metaphors of Anger, Pride and Love: a lexical approach to the study of concepts. Amsterdam/Philadelphia: John Benjamins. KoÈvecses, Z. 1990. Emotion Concepts. New York: Springer. KoÈvecses, Z. 1995. `The ``container'' metaphor of anger in English, Chinese, Japanese and Hungarian' in Z. Radman (ed.): From a Metaphorical Point of View: A Multidisciplinary Approach to the Cognitive Content of Metaphor, Berlin/New York: Walter de Gruyter. 117±47. KoÈvecses, Z. and P. Szabo. 1996. `Idioms: a view from Cognitive Semantics.' Applied Linguistics 17/3: 326±55. Lakoff, G. 1987. Women, Fire and Dangerous Things: what categories reveal about the mind. Chicago/London: University of Chicago Press Lakoff, G. and M. Johnson. 1980. Metaphors we Live by. Chicago/London: University of Chicago Press. Lazar, G. 1996. `Using figurative language to expand students' vocabulary.' ELT Journal 50/1, 43±51. Lennon, P. 1998. `Approaches to the teaching of idiomatic language'. Iral 36/1, 12±30. Lewis, M. 1993. The Lexical Approach: the state of ELT and a way forward. Hove: Language Teaching Publications. Lindner, S. 1981. `A lexico-semantic analysis of verb-particle constructions with Up and Out.' Unpublished Ph.D. dissertation, University of California San Diego. Lindstromberg, S. 1997. English Prepositions Explained. Amsterdam/Philadelphia: John Benjamins. Low, G. 1988. `On teaching metaphor'. Applied Linguistics 9/2, 125±47. McCarthy, M. 1990. Vocabulary. Oxford: Oxford University Press. Schmitt, N. and M. McCarthy. (eds.) 1997. Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press. Skehan, P. 1998. A Cognitive Approach to Language Learning. Oxford: Oxford University Press. SoÈkmen, A. J. 1997. `Current trends in
FRANK BOERS
teaching second language vocabulary' in N. Schmitt and M. McCarthy (eds.): Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press. 237±57. Swan, M. 1997. `The influence of the mother tongue on second language vocabulary
571
acquisition and use' in Schmitt, N. and M. McCarthy (eds.): Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press. 156±80. Thomson, A. J. and A. V. Martinet. 1980. A Practical English Grammar 3rd edn. Oxford: Oxford University Press.
# Oxford University Press 2000
reviews James R. Martin and Robert Veel (eds.): READING SCIENCE: CRITICAL AND FUNCTIONAL PERSPECTIVES ON DISCOURSES OF SCIENCE. Routledge, 1998.
With the evolution of science and technology, scienti®c discourse has long been the focus of interest of dierent disciplines (various professions themselves, sociology, rhetoric, history, psychology, philosophy, linguistics, education). In previously published literature in applied linguistics one might recall those highly stimulating volumes which deal with `talking science' and `writing science', revealing various aspects of scienti®c communication (Lemke 1990, Halliday and Martin 1993). This book is a collection of papers, moving the focus to the central theme of reading science. The authors, from in and around the Department of Linguistics at the University of Sydney, have been contributing in¯uential research on linguistic and educational issues related to scienti®c discourse for several decades. The present volume marks the convergence of previous research and the discussions at an invitational workshop in Australia in July 1994, with the participation of Michael Halliday, Charles Bazerman, Jay Lemke, and Frances Christie. What makes this book dierent from most previous work can be summarized as follows. First, it views reading not as `skill', `procedure', or decontextualized activity, but rather as a social practice, which encompasses a great number of interacting physical, socio-political, cognitive, aective elements in dynamically changing social contexts, which are interpreted through language. A great variety of social contexts is investigated in the book from the perspective of language use: popular science writing (Fuller); science ®ction (CrannyFrancis); industry and technology (Rose, White); scienti®c texts in medicine and physics (Lemke); school science: environmentalism (Veel); pedagogic discourse in upper primary school science: a science curriculum macrogenre (Christie); the in¯uence of science discourse on social science (Wignell) and on cognitive science (Matthiessen). Secondly, as the subtitle indicates, the discourse of science is investigated from both critical and functional perspectives. Based on the Hallidayan systemic±functional approach, scienti®c genre is also regarded as a system, where potential and actual are viewed as networks of choice. Structural descriptions are designed to relate speci®c genres to choices which were not selected but might have been. This functional approach is complemented by the critical perspective. Under the in¯uence of critical theory, another important principle is suggested through the papers: it is not enough to
REVIEWS 573
understand what form a given genre takes, it is also essential to recognize why a text/genre took the form it did in a particular context. The gap between social and textual analysis needs closing. Thirdly, special attention is paid to the fact that social contexts change; therefore understanding the role science discourse plays and could play in various changing contexts in our lives is essential as it has multiple, and highly signi®cant, consequences. The functions of science discourse are viewed in the interrelations of new technologies, rituals and values, issues of prestige and power. The power of science discourse comes from the control it aords over the material environment through technology. The more functional the discourse in social contexts, the higher the status it will attain. Language becomes ritualized when it acquires value, and it can acquire value only if it is functional. With new functions, new modes of representing knowledge also emerge. In the frame of these interrelations our attention is drawn to a wide range of important interdisciplinary factors from revealing the multimodality of scienti®c discourse to the responsibility of educational institutions. The principled approach of viewing reading as a social activity in changing contexts from functional and critical perspectives becomes a unifying force, holding the individual studies together to form a valuable volume. The book's 12 chapters by leading scholars provide a broad overview of science discourse in the community on the one hand, and science discourse for pedagogic purposes in schools on the other, from a number of diering perspectives. The editors, James Martin and Robert Veel, have arranged the papers into four parts, each with a dierent focus. Part I `Discourse on science' contains two fundamental theoretical overviews. The ®rst one `Discourse of science: Recontextualisation, genesis, intertextuality and hegemony' (James R. Martin) summarizes previous research providing the impetus for this particular assembly of papers. An especially precious rich collection of references is also included showing the previously accumulated body of knowledge. In the other study, `Emerging perspectives on the many dimensions of scienti®c discourse' (Charles Bazerman), the author overviews in¯uential achievements in rhetorical semantics, rhetorical pragmatics, grammatical semantics and rhetorical forms of symbolic actions. Four approaches to the discourse of science are considered, including Latour's actant network theory as a semiotic and rhetorical account, Greg Myers' construction of scienti®c communication as a ®eld of social negotiation, Halliday and Martin's semiotic representation of scienti®c language, and Bazerman's rhetorical forms for symbolic action. Part II bears the subtitle `Popularising science' as both papers here deal with the issues of accessibility of scienti®c language to lay persons. In `Cultivating science: Negotiating discourse in the popular texts of Stephen Jay Gould' (Gillian Fuller), the author argues that in the `translation' process making science accessible, recontextualization occurs with a variety of distinct rhetoric strategies and discursive tools applied. The author in `The ``science''
574 REVIEWS
of science ®ction: A sociocultural analysis' (Anne Cranny-Francis) suggests that it is rather ®ction than science which drives this genre as `the science of science ®ction has one major function: it establishes the constructed nature of knowledge by drawing attention to the ways in which science and/or technology determine and/or re¯ect the dominant rhetoric and ideology of a society' (p77). Part III under the subtitle `Recontextualising science' comprises papers which reveal how the reformulation of new meanings from existing contexts occurs when new contexts for scienti®c activity emerge. The interaction of dierent semiotic systems is considered in `Multiplying meaning: Visual and verbal semiotics in scienti®c text' (Jay Lemke) with a thorough demonstration that visuals do not simply `illustrate' the text, but dierent semiotic systems in modern scienti®c discourseÐcombinations of written language, diagrams, images, tables, graphs, mathematical symbols, sound images, moving imagesÐin multimedia genres are interrelated with presentational, orientational, and organizational functions, `multiplying meaning'. The dominant role of visual images in environmentalist texts is also an object of investigation in the paper `The greening of school science: Ecogenesis in secondary classrooms' (Robert Veel). A comparison is drawn between how the language of traditional science texts and environmentalist texts operates. The structure, grammatical features, patterns of ergativity in texts are analysed, and it is pointed out that the world in the latter texts is construed as a place where things just happen, rather than being caused to happen as in traditional science discourse. In her analysis of a science curriculum macrogenre, the author of `Science and apprenticeship: The pedagogic discourse' (Frances Christie) concludes that a pedagogic discourse is a consequence of the operation of a pedagogic device, whose purpose is to shape various forms of consciousness. The discourse of a school subject will be recontextualized, and it is not the logic of the science subject, but the logic of the pedagogic activity that determines its transmission to school students. Part IV `Discourses of science' begins with a chapter entitled `Things and relations: Regrammaticising experience as technical knowledge' (Michael A. K. Halliday) in the focus of which stand grammatical metaphors in scienti®c English and their systematic categorization. The author sets up a taxonomy of types of reconstrual, the categories being placed into new semantic and textual relations with one another. In the paper `Science discourse and industrial hierarchy' (David Rose), the author investigates how the potential of the language system is activated in dierent ways at dierent levels in the industrial production hierarchy. He analyses correlations of language-use characteristics in the framework of the relationship between science, education, and industry from the Australian National Board eight-level competency scales ranging from the least skilled or quali®ed production worker to the most highly trained researchers and managers. His conclusion is that the more senior levels of the hierarchy, the more metaphorical modes of expression. The lexis of scienti®c and technological discourse is analysed in
REVIEWS 575
`Extended reality, proto-nouns and the vernacular: Distinguishing the technological from the scienti®c' (Peter H. R. White). It is argued that marked patterns of dierence can be observed in the lexico-grammatical preferences of the two discourses, and these can be explained by reference to a basic dierence in the communicative purposes of their respective specialist lexis. Special attention is given to technological acronyms which are no longer simply abbreviations of longer forms but have speci®c features which makes lexicon extension possible in the domain of modern technology. The paper `Technicality and abstraction in social science' (Peter Wignell) discusses the discourse of social science, summarizing the distinctive characteristics of de®ning, classifying, and explaining in science and the humanities. The author concludes that social science uses much the same resources as scienti®c discourse in establishing a technical framework which is used for interpretation. However, the dierence can be grasped through the evolution of social science discourse in longer time frames. Like social science, cognitive science activates many of the grammatical structures of research science in order to reconstrue meanings, as explored in `Construing processes of consciousness: From the commonsense model to the uncommonsense model of cognitive science' (Christian M. I. M. Matthiessen). By examining `the mind', the metaphorical mode of construing in cognitive science, the author concludes that `within the various groups we belong to, we construe our experiences interactively, developing negotiated and collective models; our language thus constitutes our collective consciousnessÐshared also across innumerable generations' (p. 353). In sum, the great strength of the book is rooted in its interdisciplinary approach and its broad range of perspectives while discussing various aspects of reading science in a theoretically integrative framework. It should attract and stimulate a wide range of audiences as it has a great deal to oer for both academics and teachers. (Received January 2000) Reviewed by Zsuzsa KurtaÂn University of VeszpreÂm, Hungary
REFERENCES Lemke, J. L. 1990. Talking Science: Language, Learning and Values. Norwood, NJ: Ablex. Halliday, M. A .K. and J. R. Martin. 1993. Writing Science: Literacy and Discursive Power. London: The Falmer Press.
576 REVIEWS
R. K. Agnihotri, A. L. Khanna, and Itesh Sachdev (eds.): SOCIAL PSYCHOLOGICAL PERSPECTIVES ON SECOND LANGUAGE LEARNING. Sage, 1998. Vera Regan (ed.): CONTEMPORARY APPROACHES TO SECOND LANGUAGE ACQUISITION IN CONTEXT. University College Dublin Press, 1998.
In her contribution to the commemorative issue of the TESOL Quarterly on its 25th anniversary, Diane Larsen-Freeman outlined the evolution of second language acquisition (SLA) research in relation to the nature of the language acquisition process and the factors that aect language learners. She concluded that, in the future, SLA researchers attempt to combine elements from both areas in order to account for the linguistic, psychological, and social variables involved in language learning (Larsen-Freeman 1991). At the outset of a new century, the current state of SLA research, having entered what Larsen-Freeman might now consider as full adulthood, certainly appears to con®rm the above prediction. One of the most bene®cial eects of reaching this stage may be a general agreement in the SLA research community regarding the interaction of multiple mechanisms and processes in the development of the learners' linguistic system. Any possible (and convenient) tensions between lines of research on (a) universal and systematic aspects of language acquisition, and (b) individual and social variation, have moved to a dierent level, from which contemporary SLA research, while maintaining a broad perspective on language learning as a whole, attempts to provide answers to speci®c and immediate pedagogical concerns. The two books reviewed here constitute a valuable contribution to SLA research on individual and social variation, mainly because of the editors' explicit intention to present studies carried out in a variety of geographical and social contexts. (This kind of initiative may also be noted in recent collections of research on English language teacher education; see e.g., Bailey and Nunan 1996, Freeman and Richards 1998.) Social Psychological Perspectives on Second Language Learning is the fourth volume of the series Research in Applied Linguistics, which started in 1994 with the primary purpose of `giving a new orientation to applied linguistics by focusing on linguistic theory on the one hand, and socially relevant issues in multilingual societies on the other' (p. 10). In the introduction to the volume the editors provide a general description of the social psychological approach to SLA and a review of Indian studies supporting the claim that sociallyrelated factors may play a role larger than that of individual variables in SLA. The eleven papers that follow elaborate on the same position by describing research conducted in India and other geographical and pedagogical contexts. Gardner and Tremblay (Chapter 2) report on a Canadian study consisting of
REVIEWS 577
a computerized version of Gardner's (1985) Attitude/Motivation Test Battery (AMTB), and two paired-associates learning tasks (English/French, English/ Persian). They seek to substantiate their hypothesis that aective variables concerning one language are restricted to that language, but not extended to other languages. The study also indicates that (a) aective attributes in¯uence state measure of motivation (but not anxiety), and (b) state motivational attributes in¯uence SLA. In Chapter 3, Sawhney and Agnihotri's study deals with the issue of Tamil children learning Hindi as a second language (L2) in Delhi. Taking into account the presence of the target language community in this SLA context, the analysis of data from an adapted version of the AMTB pointed to the importance of dierent social psychological variables in the classroom setting (i.e. attitudes toward the teacher and the course), and with regard to the overall pro®ciency in Hindi, parental encouragement and students' academic background. Khanna, Verna, Agnihotri, and Sinha (Chapter 4) also focus on social psychological factors in an SLA environment in contact with the target language community, in this case concerning Asian adult immigrants learning English in Britain. The results of a questionnaire administered to informants from dierent parts of the country suggest that instrumental motivation appears to be stronger among adult learners than integrative motivation, due to the in¯uence of social (rather than individual) variables, such as the use of English at work and at other places. In Chapter 11, the same authors provide further observations about their results, in connection with the relationship between pro®ciency in English, and social and linguistic stereotypes among dierent groups of adult immigrant learners. The study by Khanna and Agnihotri in Chapter 5 examines variables in¯uencing the oral skills of university graduates in Delhi. The statistical analysis and interpretation of evaluations of recorded speech and learner variables (such as pro®ciency in English, personal data, and attitudes and motivation) suggest a striking relationship between oral skills and claimed used of English among friends, whereas weaker relationships with attitudes and motivation could be associated with other linguistic skills. Similarly, Agnihotri, Khanna, and Mukherjee (Chapter 6) observe a possible correlation between achievement in English tenses and speci®c social and psychological variables among Indian university students. As in the previous study, a statistical analysis of the data on learner variables and from test scores indicates the relevance of variables such as schooling, marks on the previous examination, the course being pursued, patterns of language use, and stereotypes of the English language. In Chapter 7, Sawhney examines the relation of social psychological factors and the pro®ciency level of students of German as an L2 in India (i.e. without contact with the target language community). The analysis of questionnaires, with data on social and social psychological background, motivational orientation, attitudes, and social stereotypes, reveals that variables like educational background and exposure to the L2 may be more in¯uential than attitudes and motivation to SLA. In
578 REVIEWS
general, the ®ndings of these three studies may point to the relative importance of social psychological factors according to each speci®c research context. As Sawhney points out in her conclusion, it may be that `rather than a restricted social psychological model of foreign language learning, we need a comprehensive sociolinguistic model' (p. 128). Chapter 8 comprises the most theoretically oriented contribution to the volume. Landry and Allard present a model of the integration of majority and minority languages in the education of minority group members across social contexts. Based on empirical inquiry primarily conducted in Canada and the USA, this model pertains to social/sociological and individual/psychological dimensions of bilingualism, and supports the view that in communities with a strong ethnolinguistic vitality, children may be taught entirely in the L2 and still reach a signi®cant pro®ciency in the ®rst language (L1). On the other hand, weak vitality might entail more knowledge of the L2, even if children are taught only in the L1. Chapters 9 (Giles, Harwood, Pierson, CleÂment, and Fox) and 10 (Sachdev, Elmu, and Collins) deal with stereotypes and attitudes toward speech varieties. The ®rst study compares two sets of dataÐfrom questionnaires administered to university studentsÐinvolving attitudes towards the elderly in North American culture, and in Asian and Chinese communities. Contrary to the researchers' expectations, the results suggest that stereotypes of the elderly appear to be more negative in Hong Kong, while patronizing speech seems to be less socially accepted by Californian informants. The second study focuses on attitudes held by teachers in Britain towards students' regional accents and dialects. Involving again a comparison of data from two studies carried out in London, the investigation points to an accent hierarchy in which upper-middle-class English would receive more positive evaluations than the Black London and Indian varieties, and white urban working-class English (Cockney and Birmingham) would be consistently downgraded on oral assessment criteria. Finally, Agnihotri (Chapter 12) reports on three recent studies on mixed codes and their acceptability in three contact settings (Kashmiri±English, Hindi±English, and Afrikaans±English). He emphasizes the role played by social psychological variables in the evaluation of mixed codes (e.g. positive stereotypes about languages, educational, and social background) and concludes with a reference to possible limitations in the way that current concepts of language re¯ect `what people may want to do with language in a multidimensional space' (p. 228). Contemporary Approaches to Second Language Acquisition in Social Context is a collection of papers from the Fifth Annual Conference of the European Second Language Association (EUROSLA), held in Dublin in 1995. This volume is intended to illustrate the increasing attention to social factors in SLA research, and to combine the focus of North American research on English as an L2 with the analysis of the diverse L1 and L2 spoken in Europe. In contrast to the previous title under review, the nine chapters in this volume contain insights into several perspectives on SLA and approaches to
REVIEWS 579
the collection and analysis of empirical data, in accordance with the editor's view that such diversity may be `more appropriate for an explanation of SLA than any monolithic theory' (p. 2). Pica's contribution (Chapter 1) emphasizes the relevance of `language learning through interaction' as a theme for SLA research perspectives concerned with cognitive, psycholinguistic, and social processes. The ®rst section focuses on the interaction of better-known learners' needs such as that for comprehensible input and focus on form, with other needs (e.g. restructuring and retrieval of interlanguage) that might be of equal importance when attempting to reach more dicult-to-access data in L2. Next, the discussion deals with (a) negotiation of meaning, in relation to other cognitive and psycholinguistic constructs, (b) collaborative dialogue, as a feature allowing scaolding, completion, and production of modi®ed input, and (c) instructional intervention, in terms of the lack of balance between current measures of comprehension and communicative ¯uency, on the one hand, and versus production and linguistic accuracy on the other. The following two chapters report on studies within the interactionalist perspective. In Chapter 2, Langman examines the use of a pragmatic marker of understanding as a communication strategy in the context of Chinese speakers of Hungarian in Budapest. The data were collected through a number of interviews, which were transcribed and analysed using the CHILDES (Child Language Data Exchange System). Hickey (Chapter 3) combines several research techniquesÐpro®les, questionnaires, and multivariate statistical analysisÐin order to determine what factors may contribute to successful L2 learning of Irish among kindergarten children in Ireland. The ®rst study claims that communication strategies may involve linguistic pro®ciency as well as the strategies and linguistic repertoires of all the interlocutors. The second suggests that the provision of input, as a major causal factor in language acquisition, should be accompanied by diverse teaching techniques aimed to provide sucient opportunities for interaction. Chapters 4 and 5 look at the concept of reference in L2 speech. First, Lambert compares data concerning form±function relations from a retelling task administered to American native speakers, French native speakers, and American advanced learners of French. The results seem to con®rm that particular linguistic preferences and structures may be strongly in¯uenced by the way in which adult learners of L2 learned how to view and report events and experiences in their L1. Next, Bos describes two studies (from a larger longitudinal study) with two age groups of Moroccan children living in the Netherlands. In order to examine the children's comprehension of anaphoric references and relative clauses, and their topic continuity (maintaining reference to a character of a narrative), semi-spontaneous speech data were collected through a combination of retelling and elicitation techniques, and transcribed following the CHILDES and CHAT (Codes for the Human Analysis of Transcripts). The analysis of time spent in the native speaker community is addressed
580 REVIEWS
from two complementary perspectives. Freed (Chapter 6) compares two groups of American-English learners of French (one went abroad, the other remained on campus) and describes the perceptions of L2 native speakers of the learners' speech. Although other collection methods were used, here the researcher gives account of (a) the oral pro®ciency interviews, with a slight advantage regarding progress made for the students who went abroad, (b) the judges' evaluation of perceived ¯uency, from which a signi®cant dierence in the speech of less advanced students returning from abroad was observed, and (c) the speci®c linguistic dierences between the two groups. To investigate interstylistic and interindividual variation in L2 learners' speech rate, Dewaele (Chapter 7) focuses on Dutch L1 speakers learning French in a Belgian university. Samples of `informal' and `formal' speech were gathered (from unstructured interviews and presentations, respectively), along with measures of learner extroversion and speech rate. The variables that aect speech rate appear to be the situation (speakers slow down in more formal contexts), sex (female speakers may be more ¯uent in both contexts), and his or her degree of extroversion (extroverts perform faster). The ®nal two chapters are contributions to variationist work in SLA. Dittmar, Spolski, and Walters (Chapter 8) propose a framework for the study of language acquisition in use, combining concepts and qualitative/quantitative research techniques from sociology (external social context), linguistics and sociolinguistics (language acquisition and language change), and social psychology (individual identity and attitudes). This framework is illustrated with the description of an ongoing comparison of immigrants in several target communities in Israel. The tentative ®ndings point to a noticeable change of orientation in the grammaticalization process, a change mainly related to innate/cognitive processes in early stages and driven by social forces as it advances within any given community. In Chapter 9, Poplack and BudzhakJones examine language use and linguistic results of language contact in a small group of informants living in Ottawa, Canada, who are bilingual in Ukrainian and English at dierent levels. A corpus of speech from a series of unstructured interviews, analysed through a stepwise multiple regression procedure, outlines the variables aecting the treatment of both languages by ®rst generation speakers (prevailing use of Ukrainian), and second generation speakers (prevailing use of English). The results emphasize that studies of bilingual discourse should take into account the distinction between native and non-native speakers, and the related classi®cation of speci®c data as instances of code-switch or lexical borrowing. These two volumes would certainly provide any interested reader with a remarkable amount of information concerning individual and social factors in various SLA contexts. In addition, the editors include theoretical overviews on social and psychological dimensions in bilingual education (Landry and Allard), the interactionist perspective in SLA (Pica), and a proposed research framework for variationist analysis (Dittmar, Spolski, and Walters). However, the range of research sites and to a certain extent, theoretical approaches, does
REVIEWS 581
not appear to be as diverse when it comes to the implementation and/or combination of dierent methods for the collection and analysis of data. In this respect, the in¯uence of a speci®c theoretical approach to researching SLA in context (as in the volume of Agnihotri, Khanna, and Sachdev) might limit the methodological choices that researchers could make, and may aect their ®ndings. Regan's collection includes accounts of a larger range of research techniques, often described in more detail than in the previous compilation. But in this case my concern is the limited number of theoretical perspectives on SLA in context in Regan's volume, especially considering the complexity of this ®eld of inquiry. This limitation could be due to the small number of contributors, and to the fact that most chapters report on research concerning speci®c issues or variables. Perhaps a more comprehensive volume containing a lengthier and more balanced selection of empirical and theoretical studies might provide a more general portrait of SLA in social context. (Received March 2000) Reviewed by Manel Lacorte University of Maryland-College Park
REFERENCES Bailey, K. M. and D. Nunan. (eds.) 1996. Voices from the Language Classroom. Cambridge: Cambridge University Press. Freeman, D. and J. C. Richards. (eds.) 1998. Teacher Learning in Language Teaching. Cambridge: Cambridge University Press. Gardner, R. C. 1985. Social Psychology and Second Language Learning: The Role of Attitudes and Motivation. London: Edward Arnold. Larsen-Freeman, D. 1991. `Second language acquisition research: Staking out the territory.' TESOL Quarterly 25/2: 315±50.
Gunter R. Lorenz: ADJECTIVE INTENSIFICATIONÐLEARNERS VERSUS NATIVE SPEAKERS. A CORPUS STUDY OF ARGUMENTATIVE WRITING. (Language and Computers: Studies in Practical Linguistics 27.) Rodopi, Amsterdam 1999.
Adjective Intensi®cationÐLearners versus Native Speakers: A Corpus Study of Argumentative Writing is an interesting study of various types of intensi®ers of adjectives in English. The study is particularly interesting with regard to pedagogy and corpus methodology, but it has also descriptive implications, which should be of great heuristic value. Lorenz' study is the ®rst monograph that makes use of material from the International Corpus of Learner English (ICLE), consisting of essays written by university students, both advanced learners and native speakers, British and American. The ICLE data are
582 REVIEWS
supplemented by material from younger speakers, i.e. both learners and native speakers below university level. Three main goals are addressed in this study. They are: 1 2 3
to implement corpus methodology in the domain of learner language; to give a functional account of intensi®cation of adjectives in written (argumentative) English by native and non-native speakers; to explore native and non-native usage and the notion of idiomaticity.
In order to implement his corpus methodology, Lorenz explores four corpora of argumentative writing. Two corpora are sampled from advanced German learners of English and two from native speakers of English. Corpora
Age
Size (number of words)
German teenagers German undergraduates (ICLE) British teenagers British undergraduates (ICLE)
16±18 20±25 15±18 19±23
130,730 72,031 68,549 92,008
The text-functional category of argumentation is kept constant, while native and non-native background and age are variables under scrutiny. Lorenz brings up the problem of how large a corpus needs to be in order to be a reliable sample for a study of collocational patterns. Although the corpus is comparatively small, Lorenz claims that it is large enough for a lexicogrammatical study of this kind, since he neither attempts at determining frequencies, nor all possible combinations of intensi®ers and adjectives. The focus of this inquiry is instead on more general patterns of collocability and corpus methodology. Throughout the book Lorenz makes use of corpora and statistics in linguistics in a sensible and insightful way. His judgements and conclusions are well argued and well balanced, and the study is characterized by creativity and originality. Lorenz argues that if we assume that interlanguage is systematic, it should be possible to investigate learner language in corpora in the same way as naturally occurring native language. He starts out with a number of hypotheses, which provide the basic rationale for large-scale investigations of learners' written production. First, machine-readable learner corpora are mainly useful for large-scale studies that can help the analyst detect `over-indulgence' and `under-representation'. Furthermore, linguistic choices can be described on a scale of `commonness' ranging from deviant to idiomatic. The impression of unidiomatic as well as idiomatic language is not only a matter of degree, but it is also cumulative and should therefore be quanti®ed in a large corpus of learner language in order to uncover patterns
REVIEWS 583
on the cline from deviant to idiomatic, rather than to focus on errors. Learner corpora are best suited for studying advanced learners, since idiomaticity is part of the ®ne-tuning. Finally, Lorenz also presents a heuristic hypothesis; namely that idiomaticity can be improved by making covert patterns overt, since learners resort to conscious knowledge, in particular in writing. Lorenz describes the combinatorial potential of the functional paradigm of intensi®ers and adjectives and identi®es the conceptual mechanisms of its dynamic and innovative character. The line of inquiry is strictly functional; i.e. it goes from function to form. He wisely includes a wide range of modi®ers in the scope of his study. First he divides the intensi®ers into closed-class items, such as very, almost, and quite, and open-class items, such as reasonably, badly, and crucially. The closed-class intensi®ers were found to be overused by the learners. The open-class category was further divided into ®ve subcategories based on by what capacity the adverb functions as an intensi®er. Open-class intensi®ers were found to operate along ®ve dierent semantic dimensions: 1 2 3 4 5
scalar (completely, widely, largely, slightly); modal (clearly, certainly, possibly, truly); evaluative (awfully, ludicrously, suciently, suitably); comparative (especially, comparatively, particularly, extraordinarily); semantic feature copying (loosely (structured), easily (possible), universally (equal), darkly (tanned) ).
According to Lorenz the above ®ve categories have proved to be of high conceptual value. However, it is not entirely clear what Lorenz means by conceptual value. He probably means that they are relatively strong on propositional meaning in contrast to the closed-class items. The scalar set includes ly-adverbs, which have no semantic extension beyond that of scaling an adjectival property. They are either scaling by denotation or grammaticalized to this function. They are thus similar to closed-class items in some respect. Scalar modi®ers are slightly over-used by German learners as compared to native speakers. On the surface, it looks as if German learners also overuse modal intensi®ers, and quite substantially so. However, it was found that the German learners use really to an extent that would normally only be expected in casual speech, while the native speakers rarely use it in argumentative writing. In fact, this learner overuse of really over-compensates what would otherwise be an underuse of modal intensi®ers. Evaluative intensi®ers are underused by the learners. On the whole, evaluatives are not used very much by either of the two populations. The comparatives comprise an even smaller category than the evaluatives, and are characterized by learner underuse. Finally, the semantic feature copying category consists of items which achieve intensi®cation by duplicating part of the meaning of the adjective. Lorenz says that this category is perceived as particularly idiomatic. The learners' de®cits in this category contribute to the impression of `lack of originality' in the most striking way.
584 REVIEWS
These results appear much more interesting in terms of aspects of learner usage versus native usage than as categories of the same functional paradigm. There is more to be desired with respect to the theoretical background of the categorization. For instance, why are terribly and awfully included among the evaluatives and not among the scalar modi®ers or even among closed-class intensi®ers? Why are scalar modi®ers said not to be feature copying? To me terribly and awfully lack evaluative features (in Lorenz' sense). They are just intensi®ers without extension beyond scaling (terribly nice, terribly bad). Scalar modi®ers are de®nitely feature copying. They are not constrained by features in the content domain, but by the mode of construal, i.e. whether the adjective is perceived in terms of a scale or a limit. Lorenz makes use of both quantitative and qualitative methods of analysis. He makes use of native-speaker informants for judgements of acceptability, and he uses Mutual Information (MI), which is an arithmetic operation to express idiomaticity. MI is a value used to identify signi®cant collocations from large corpora (p. 174). Lorenz presents his results in a clever way by applying several frequency ®lters on the various MI-rankings. The lower the frequency ®lter is the greater the perceived degree of idiomaticity of the expressions. The highest ®lter used for potentially signi®cant combinations is 5, i.e. combinations that occur ®ve times or more. Naturally, such a high frequency ®lter would not produce very interesting combinations. The openclass booster very represented more than half of them. The ®lter was gradually lowered down to un®ltered combinations, where the correlation with perceived idiomaticity is much greater. Combinations such as immensely troubled, boringly sti, and darkly tanned were found here. The outcome of Lorenz' study is that the semantic copying category represents the more idiomatic type as opposed to closed-class and scalar modi®ers. Lorenz relates his ®ndings to Sinclair's open-choice principle and idiom principle. He concludes that the learners make more use of the open-choice principle, i.e. the highly delexicalized items, while the idiom principle is said to be at work mainly in the semantic feature copying category (p. 215). This statement seems somewhat distorted. Unfortunately, Lorenz does not discuss the open-choice principle and the idiom principle in a penetrating and principled way, nor does he explain exactly what he means by idiomaticity. It appears as if he equates idiomaticity and originality. As a result, the correlation between the open-choice principle and the idiom principle vis-aÁ-vis idiomaticity remains unclear. It seems to me that his ®ndings would make an excellent basis for a more sophisticated analysis of the open-choice principle and the idiom principle and, for that matter, what lies between these two extremes. The study would pro®t from some further discussion, both of the combinatorial cline per se and from the point of view of the implications of the theoretical complexity associated with language learning as a cognitive process. Apart from the ®ndings that the learners go for safe bets, i.e. that they are much more inclined to use closed-class, `all-purpose' intensi®ers and highly delexicalized scalar open-class item, the real surprise according to Lorenz was
REVIEWS 585
that the learners make far more use of adjective intensi®cation than native speakers do, both in absolute ®gures and in relation to the number of adjectives used. Lorenz posits a number of hypotheses to explain this, the most plausible being that native speakers can do without them, since their active vocabulary is much larger and more variable. Learners make use of common core adjectives that need re®nement, while native speakers might use other lexical means of achieving the same expressivity. The numerical pattern of learner overuse points to information overload, one of the main stylistic weaknesses of non-native writing according to Lorenz. To summarize, the topic under investigation is not only original and well researched, the ®ndings are also well presented and the language is highly idiomatic. Lorenz has managed to show that one of the strengths of corpus linguistics lies in its potential for highlighting patterns that would otherwise be hidden from the naked eye. Lorenz has implemented a methodology to bring out systematic irregularities, which in their individual occurrences appear just marginally inappropriate or accidental. Apart from the minor drawback that the book lacks an index, it is reader-friendly and clear. After each chapter Lorenz makes a summary, and at the end of the book there are eleven appendices, which supply the reader with the data drawn from the corpora. The book is a good read. I sincerely recommend it. (Received March 2000) Reviewed by Carita Paradis Lund University, Sweden
John M. Swales: OTHER FLOORS, OTHER VOICES: A TEXTOGRAPHY OF A SMALL UNIVERSITY BUILDING. Lawrence Erlbaum 1998.
Oh! Blessed rage for order, pale Ramon, The maker's rage to order words of the sea, Words of the fragrant portals, dimly-starred, And of ourselves and of our origins, In ghostlier demarcations, keener sounds.
So ends Wallace Stevens's great poem `The Idea of Order at Key West', which speaks of the human need to ®nd order in the world. The task of the interpreterÐthe makerÐis, Stevens insists, to listen to the voices of those who share the world, the `others'. In the poem the poet creates order not by observing the sea at Key West but by listening to the singer: She sang beyond the genius of the sea ..it was she and not the sea we heard.
586 REVIEWS
This `blessed rage for order' is what motivates both the arts and the sciences, all intellectual endeavour. But in creating order, whose order is it? That of the maker or that of the others? This dilemma, this postmodern, post-colonial, post Enlightenment challenge to hegemony, has stymied con®dence in the academy, particularly in the social sciences. And within the social sciences anthropology has been hardest hit. As Cliord Geertz tells us: `The ®rst index entry nowadays in books surveying the ®eld is often ``anthropology, crisis of . . .'' ' (Geertz 1995: 97±8). But for Geertz there is a solution to the dilemma if `anthropologists recognise that the anthropological vocation (is) in important respects a literary one' (Geetz 1988: 142). Geertz's use of the term `literary' to refer to the anthropologist's enterprise, parallels the role Stevens attributes to the poet: both are being creative, interpreting, `making'. What is needed, Geertz continues, is `this capacity to persuade readers. That what they are reading is an authentic account by someone personally acquainted with how life proceeds in some place, at some time, among some group' (Geertz 1988: 143±4). He is con®dent that anthropology has a future which will `enlarge the possibility of intelligible discourse between people quite dierent from one another in interest, outlook, wealth and power' (Geertz 1988: 147). But, Geertz points out, this is not enough: `doing so does not relieve one of the burden of authorship; it deepens it' (Geertz 1988: 140). The anthropologist may have left his/her imperial mission behind, but he/ she remains the interpreter of the `other' whose task it is to bring the `there' `here', to impose order which may no longer be hegemonic but is still order. That responsibility of authorship is the poet's and the anthropologist's, it is the literary vocation. The order that it describes, while imposed, through authorship, by the maker, is an order that emerges by listening to the singers telling us what they are like, not by our telling them what they should be like. In expanding beyond his earlier work on genre (Swales 1990) to the discourse communities he describes in such loving detail in Other Floors Other Voices (OFOV), it is to Geertz that Swales looks for inspiration, as a model both for style and purpose. Swales praises Geertz's `magic prose' (p. 22) and refers to his own `entrancement' with Geertz's essays (p. 179). No author (other than Swales himself) is cited in OFOV more than Geertz. All of which might lead us to regard OFOV as a type of Geertzian anthropology. Is this what it is? The book's sub-title is: `A textography of a small university building'. The building, North University Building (NUBS), situated on the campus of the University of Michigan at Ann Arbor, contains three units, the Computing Resources Site (CRS), the University Herbarium and the English Language Institute (ELI). Swales points out that `none of the building's three units is a typical academic unit, such as a Department or a Program. All three are involved with services' (p. 3). He is concerned to establish their dierence from one another and at the same time to demonstrate that they are all valid and recognizable parts of the academic enterprise. To make his point, Swales oers a textography. By textography he means
REVIEWS 587
`something more than a disembodied textual or discoursal analysis, but something less than a full ethnographic account' (p. 1). Ethnography is unacceptable to Swales as a label for his enterprise. He prefers the label `textography', which seems to refer to the literate activities that sustain and explain institutions. The study is site bound, context dependent, oering a `study of particularities and communality as seen through the lens of written discourse' (p. 1). By the end of the book Swales has decided that his textography is concerned with a type of discourse community, what may be termed a `place discourse community' (PDC). This is reminiscent of earlier attempts to develop a branch of linguistics concerned with systematic language uses, registers, genres, and indeed the discourse of place, but the suggested term `institutional linguistics' (Halliday et al. 1964) has not been widely used. Swales provides four main criteria for a place discourse community (p. 204), which we can summarize as follows: 1 2 3 4
It is a group of people who regularly work together (if not always or all the time in the same place). It has evolved a range of spoken and written genres. It has developed a consensus regarding its work practices. It has a sense of its own history.
After the Introduction the book falls into three parts. In the ®rst (Chapter 2) Swales analyses `the quotidian discourse-related activities that occur and recur on the three ¯oors of this building. I have been selective here choosing certain more or less regular processes that seem emblematic of the way things generally get done on each ¯oor. These are ``the de®ned rhythms of work'' that orchestrate the roles of text and task and are, in turn orchestrated by them' (p. 28). On each ¯oor he chooses one activity that is a more or less regular process. In the CRS he examines `the roles of the consultants as purveyors of technical advice and customer help'. In the Herbarium he describes the loans of vascular plants to other institutions. And in the ELI he selects the Testing and Certi®cation Division. In the second part, Swales deals with the written discourses of individuals, ®rst (Chapter 3) in the Herbarium and second in the ELI (Chapter 4). The CRS is not included in this part because `most of the textual production associated with computer research actually takes place elsewhere' (p. 25). In the third part (Chapter 5) Swales re¯ects on his textographic analysis and considers how robust is what he calls the `troubled concept of discourse community' (p. 2). To an extent the study as a whole operates as an attempt to validate this concept. As Swales explains `the NUBS study can be viewed as a kind of test case. In eect if the concept of discourse community does not, in some sense, survive its encounter with NUBS then it probably is of diminished utility and of somewhat exceptional occurrence' (p. 23). Swales was surely lucky in his building: a Computing Resource Site, a University Herbarium, an English Language Institute (with its own Testing
588 REVIEWS
Division): what units in close proximity could be more diverse? Surely these must provide distinctive voices. Do they? Swales explains the development in his own thinking of the concept of the discourse community. Rather than taking membership or role as primary he recounts how he came to the notion of a body of content `which gets variously shaped by convention, communicative purpose and audience design' (p. 22). This is the approach via culture, and what it allows Swales to do is to examine and then demonstrate how all the participants in each of the three units are collectively engaged in and share the same cultural endeavour, even though of course their own training and academic status may be very dierent. In the Herbarium non-academic as well as academic sta acquire the culture of craft-knowledge particular to their Unit, they do this through what Lave and Wenger (1991) term `legitimate peripheral participation' (p. 51). In the ELI he welcomes the encouragement given to developing `a vibrant interpersonal and professional life (to) exist within a community of practice engaged on what most people would assume to be a series of humdrum pursuits' (p. 76). And in their common culture all those who participate belong to their community of scholars, whether engaged in research or service in the Herbarium, whether in research or practice in the ELI. In his last chapter (`Re¯ections') Swales concludes that he has distinguished `three dierent voices' in NUBS but that only one unit, the Herbarium, quali®es unreservedly as a discourse community, or as he prefers `a place discourse community'. The claim of the ELI is slightly weaker than that of the Herbarium, but is still valid. The CRS, however, does not qualify, largely it seems because it is very much an out-station of the main computing centre in the university and because of the paucity of its hardcopy written texts; e-mail is not considered as oering hardcopy data. In spite of his own membership of the ELI and his own inclusion as a text biography in the account of that Unit, it seems clear that the Unit that really enthuses Swales is the Herbarium. This is understandable since Swales's own approach to texts is that of a taxonomist: he reminds us in passing of his own long-standing interests in stamp collecting (p. 49), in coin collecting (p. 177) and in bird-watching (p. 112). And therefore it is not surprising that his biographer (Tony Dudley-Evans), who takes on the role of Swales-asinterviewer when Swales himself is being interviewed, should make the following comment on Swales's oeuvre (for example Swales 1985, 1990) `you haven't written a theoretical justi®cation for ESP. Nor do you base Genre Analysis on a linguistic or sociological theory' (p. 182). Does OFOV succeed in demonstrating that the three units in NUBS have distinctive voices? Yes, it does. The thick description of the work and the people involved in those Units is an excellent example of that kind of ethnography which Geertz sees as helping solve the crisis in anthropology. For OFOV is, after all, an ethnography, even though Swales dislikes the term. And so is the conclusion that two of the three can be regarded as discourse communities well made? Here I am less con®dent.
REVIEWS 589
In OFOV Swales provides a Geertzian account by `someone personally acquainted with how life proceeds in some place, at some time, among some group'. He meets the members on the three ¯oors, he makes possible `intelligible discourse between people quite dierent from one another', but he is still intentÐquite properlyÐon imposing order. Is it his order or is it the order of the singers? And to what extent is that order, theirs or his, one that is created, as he would propose, by means of their textsÐthis is after all a `textology'? In my view his success is partial. He succeeds admirably for the Herbarium `the Herbarium easily quali®es for PDC status' (p. 205); but is not quite as successful for the ELI and not at all for the CRS. Hardly surprising when we look back at the criteria for a PDC. Both the Herbarium and the ELI have been in existence for a long period with the same core members of sta, both are somewhat peripheral to their Faculties, both concerned largely with activities beyond the University. Unlike the CRS, the Herbarium and the ELI have had the timeÐand the needÐto develop their own speci®c cultures. But there is a further reason. The idea of order in the Herbarium comes not primarily from an analysis of its texts but from its commitment to a taxonomic approach, its concern that science should not be thrown into confusion. The Herbarium does of course use texts taxonomically, but it is the principle of taxonomy that is prior. Swales has attempted to impose a similar order throughout all three discourse communities. It was a brave attempt but one that was doomed because neither the CRS nor the ELI is taxonomic in principle and their uses of texts do not make them so. Is it possible that the concept of place discourse community is itself `of diminished utility' (p. 23), one that Swales over-reached himself in imposing? If so, the reason for his chief success with the Herbarium would be their commitment to taxonomy, or Swales's certainty that that was the case. But the songs sung in the ELI and the CRS were dierent songs and it was less easy in those units to impose a taxonomic solution. For applied linguistics, textography oers an innovative and worthwhile link between studies of language and studies of culture. Swales is to be congratulated for showing the way and it is to be hoped that he will pursue his search for order through this new research methodology. If he does so, he will need to subject himself even more to what Levi-Strauss called the ethnographer's `chronic uprootedness'. It may therefore be sensible to leave behind his bird-watching, his stamp and coin collecting, and indeed his own curriculum vitae. It was probably a mistake to include himself as the ELI's main man: too taxonomic, too much `Here'. (Received April 2000) Alan Davies Edinburgh University
590 REVIEWS
REFERENCES Geertz, C. 1988. Works and Lives: The Anthropologist as Author. London: Polity Press. Geertz, C. 1995. After the Fact: Two Countries, Four Decades, One Anthropologist. Cambridge, MA: Harvard University Press. Halliday. M. A. K., A. McIntosh, and P. D. Strevens. 1964. The Linguistic Sciences and Language Teaching. London: Longman. Lave, J. and E. Wenger. 1991. Situated Learning: Legitimate Peripheral Participation. Cambridge: Cambridge University Press. Stevens W. 1947. Poems. New York: Random House. Swales, J. M. 1985. Episodes in ESP: A Source and Reference Book on the Development of English for Science and Technology. Oxford: Pergamon. Swales, J. M. 1990. Genre Analysis: English in academic and research settings. Cambridge: Cambridge University Press.