1

In G. Cook & B. Seidlhofer (Eds.), Principle and Practice in Applied Linguistics: Studies in honour of H.G. Widdowson . Oxford: Oxford University Press, 1995: pp. 107-124. Situational context, variation, and second language acquisition theory* Elaine Tarone University of Minnesota Guo-qiang Liu Deakin University Abstract This paper presents a rationale for examining second-language acquisition (SLA) using a variationist perspective, and demonstrates how this perspective can help us to understand and explain the process of acquisition of a second language. Data from Liu (1991; under review) show how a learner's participation in different kinds of social interactions resulted in variable production of IL forms at any given point in time, with a consequent alteration of the rate and route of acquisition of English question forms. ************************************************************* Systematic interlanguage variation occurs when a learner, at a given point in time,. produces different variants of a particular IL form under different social conditions (e.g. with different interlocutors, or in different physical locations) (cf. Tarone 1979, 1988). A learner may seem to have mastered a given target language (TL) form in one social context, yet in another social context systematically produce a quite different (and possibly inaccurate) variant of that form. Should second-language acquisition theories account for such variation in interlanguage performance? Variationist and non-variationist researchers give different answers to that question. Non-variationist theoreticians (e.g. White 1989; Gregg, 1990; 1993) focus upon internal mental processes and model an abstract (non-contextualized) competence, discounting the importance of variation

An earlier version of this paper was presented at the 22nd University of Wisconsin Milwaukee Linguistics Symposium, on Second-Language Acquisition Theory and Pedagogy, October 9, 1993. An earlier version, focusing on pedagogical implications, appears in a volume containing the proceedings of that conference. We are grateful to Fred Eckman, for his encouragement and advice on previous drafts of this paper. Our thanks go also to several anonymous reviewers of an earlier draft. Errors, inconsistencies, lapses and omissions are of course our own.

2
in learner performance; thus, they prefer to abstract away from variable data in explaining the acquisition of a second language. For such investigators, contextual variation is a performance factor, and so by Gregg (1990), for example,

definition is irrelevant to the description of the learner's idealized competence.

argues that interlanguage variation should not be the concern of second-language acquisition theorists, who should concern themselves solely with idealized competence and not with performance. In contrast, Tarone (1983, 1990) and other variationists (e.g. Dickerson 1975; Ellis 1985, 1987; Young 1991) focus upon external interactions and L2 learner performance in a variety of social contexts, and believe interlanguage variation across those contexts to be importantly related to change in learners' IL knowledge over time. Variation, from this perspective, is a source of information about the way in which interaction in different social contexts can influence both interlanguage use AND overall interlanguage development. Tarone (1983, 1988, 1990) has argued that it is important for any second-language acquisition theory to describe and explain why it is that interlanguage performance varies systematically from one social context to another, and to relate this variation in performance to the development of the interlanguage system. How can this difference of opinion be resolved? Eckman (1991 & forthcoming) suggests that variationists need to produce data which show, in precise terms, why the study of second-language acquisition must include an analysis of interlanguage variation. Can data be found which show that IL variation affects or reflects acquisition processes? Can careful studies of variation in IL performance produce data which shed light on the cognitive process of second-language acquisition? In this paper, we will argue that an adequate theory of second-language acquisition must include an account of IL variation, and give an example of the kind of data which a variationist approach can produce, data which show that a learner's participation in different kinds of interactions can differentially affect the rate and route of the acquisition process. An Adequate Theory of SLA Needs to Account for IL Variation We will begin by arguing, as has Tarone before (1983, 1988), and as many others have argued far more persuasively (cf. Hatch 1978; Wong Fillmore 1976; Selinker and Douglas, 1985; Preston 1989; Sato 1990) that in this field we have tended to err by divorcing the study of the internal development of an IL

3
grammar from the study of the (external) social contexts in which the learner develops this grammar. The social context of the learner is typically viewed as unrelated to the internal cognitive process of L2 acquisition, and so is unworthy of comment in many SLA studies, particularly those in the UG tradition (see, e.g., Rutherford 1984; White 1989; papers in Gass & Schachter 1989, and in Flynn & O'Neil 1990). As a result, our understanding of the importance of research on IL variation has been restricted. Historically we have understood "variationist" research to be focused solely upon the way in which different social variables cause patterned surface-level shifts in IL forms in performance (e.g. Dickerson 1975; Young 1991). Such shifts have typically been viewed by many writers in SLA (e.g. Gregg 1990) as temporary, ephemeral, and therefore unimportant to the diachronic change of the IL system: merely performance factors, but not factors important to the development of competence. We have not gone on to ask or to find out whether a learner's participation in different kinds of social interactions can result in different patterns in the longitudinal development of IL grammars. We have not tied together several strands of research which could have been related: research on IL variation (e.g. Dickerson 1975; Beebe l980); research on the role of interaction and input in SLA (e.g. Pica & Doughty, 1985, Long 1985); research on discourse domains (e.g. Selinker and Douglas, 1985) and research on the longitudinal development of grammatical forms (e.g. Schumann 1978, Pienemann and Johnston 1987.) Work on variation is related to work on interaction and input, on discourse domains, and on the development of IL forms. Exactly how can we show that synchronic variation in interlanguage performance is related to diachronic second-language acquisition? Are there data which clearly show that the study of IL variation in different interactional contexts can provide us with important insights into the development of interlanguage over time? Two Studies Focusing on Interlanguage Variation 1. Markee Study One could take the position that data alone will not solve the problem. If the essential difference between variationists and non-variationists is simply one of focus, then data could be interpreted in different

4
ways depending upon the focus of the interpreter; the choice would be simply a matter of style, or convenience. For example, variationists might see in the example of the acquisition/learning of the word "coral" in Markee (1991 & forthcoming), an important piece of data which shows the way in which a series of interactions in English (L2) cause a learner to variably approximate the TL norm. In this study, Markee observed a woman in an English as a second language (ESL) class as she participated in a series of group discussions which led up to her presentation in front of the class on the importance of coral reefs in the ecology of the ocean. The learner, in her attempt to internalize the lexical item "coral", variably approximated the TL norm by gradually eliciting pieces of the definition of this term in a succession of interactions with several interlocutors. At any one point in time in the conversation, the learner had a tentative and variable notion of what "coral" was (in Tarone's terms, her knowledge itself was unstable in addition to not being equivalent to the TL norm). The learner sought and found additional information about "coral" in her interactions with various interlocutors in small group discussions, resolved her uncertainty about the meaning of the term, gradually added more information about the term in additional interactions, and finally emerged in her formal presentation with the ability to give a fairly target-like definition of the term. A non-variationist might argue that what really matters in this case is the internal processing done by the learner in building up a definition of the term, and that the variable ways she uses the term in her interactions with others are irrelevant, or at best, highly subsidiary, to those internal processes. A variationist, on the other hand, might argue that the learner's "external" negotiated interactions over time with several interlocutors provide the driving force for her variable build up of an understanding of the term; her variable productions and her interlocutors' reactions to them are central to the acquisition process and in some sense trigger her internal processing. Clearly, a variationist and a non-variationist could interpret this bit of data differently; in this study there is nothing to persuade us that a study of external interactions engaged in by this learner will shed any light upon our understanding of this learner's internal processing.

5
What is needed are data which can show that the study of interlanguage variation can provide insight, obtainable in no other way, into the mechanisms of second-language acquisition. In other words, data are needed which show that the external processes studied by variationists can affect in non-trivial ways the internal processes studied by non-variationists. The Kind of Data Needed to Show Impact of Variation on SLA The kind of data we need at this point in time, we would argue, can only be produced by longitudinal case studies in which records have been kept of a learner's output in clearly different social interactions and the changes in that output in those situations over a period of some time. That is, if we are to show that synchronic change across situations is related to diachronic change, then we must gather data, individual by individual, longitudinally across an identical set of situations. We are hampered by the fact that, to our knowledge, such studies are hard to find, possibly even nonexistent until recently. Most studies on second-language acquisition are, oddly enough, not longitudinal: they do not follow a learner or learners over a period of time. This is because longitudinal studies take a great deal of the researcher's time and effort in relation to publications produced and strength of claims made. As a result, longitudinal studies are rarely done by anyone other than graduate students (the only recent exception to this generalization of which we are aware is Schmidt and Frota, 1986). Further, most studies in SLA do not focus upon individuals; the emphasis has been upon studies which sample data taken in structured situations from large numbers of subjects, so that the results can be claimed to be representative of some larger population. Markee (1991 and forthcoming) has clearly pointed out the limitations of large-scale studies of this type, and argued persuasively that certain kinds of evidence in SLA can ONLY be obtained by means of the careful observation of individuals in social context. Evidence of the value of IL variation for SLA, we would argue, is that kind of evidence, gathered longitudinally. Even when longitudinal case studies have been undertaken, data have not been gathered in distinctly different social contexts and analyzed to determine whether the IL developed differentially in those different contexts. In Schumann (1978), for example, data gathered over time in a variety of different social contexts were pooled and analyzed together. Thus, even when longitudinal case studies have been done, because of the foci of these studies, the data obtained have not been analyzed in light of the research question posed above:

6
how does IL variation in different interactional contexts relate to the development of interlanguage over time? We must recognize that the sort of data we need is extremely difficult to come by at the present time. Nevertheless, one example of such data does exist, and that study supports the view that different kinds of social interactions can differentially activate internal acquisition processes. We will here attempt to describe that study in sufficient detail to demonstrate the value of the data generated in a study of this sort; an exhaustive description of the study is provided elsewhere. 1 2. A Study of Bob's Interlanguage Guo-qiang Liu (1991; under review) conducted a 26-month longitudinal study of a Chinese boy named "Bob", beginning when Bob was almost five years old and ending when he was almost seven. Liu observed Bob speaking English in Australia in four interactional contexts: (1) first, interacting with pre-school peers and supervisory staff, and later (2) interacting with his teachers in his primary school classroom, (3) interacting with primary school peers in the same classroom, and (4) interacting with the researcher. Over the 26-month time period, 83 sessions were recorded, most of them on videotape. Sessions 23 to 83 took place during the period when Bob was in school. Each of 37 sessions recorded in the school classroom lasted about an hour; the 23 sessions with the researcher over the same time period lasted about 45 minutes. The researcher was present at all sessions, an observer in the first three contexts observed, and a participant observer in the fourth. All the data were transcribed and analyzed according to several general criteria, including number of utterances produced by Bob, complexity of utterances produced, proportion of those utterances which were initiations as opposed to responses, functions those utterances fulfilled (cf. Nicholas 1987), and interrogative structures produced. First, let us examine the general nature of the three different types of social interaction observed over the period of time when Bob was in primary school, and the general effect of those interactions upon

For a more detailed account of that study, we refer the reader to the dissertation (Liu, 1991) and a paper based upon it (Liu, under review). At this point in time, Liu has only begun to analyze his data in any depth; we can look for a deeper and more detailed analysis of the interaction between social forces and Bob's acquisition of specific grammatical forms in future papers.
1

7
Bob's IL production during that time period. We will then report Liu's analysis of the development over time of question forms in Bob's interlanguage, and his evidence that the rate and route of Bob's acquisition of question forms was differentially affected by his participation in those different social interactions. A General Description of the Social Interactions and Bob's Production of IL in Those Interactions Liu identifies three interactional contexts in which Bob engaged; he defines these contexts in terms of Bob's role relationships with three different types of interlocutors: classroom teachers (a childstudent to adult-teacher role relationship), peers in the classroom (a child to child-friend role relationship), and the researcher (a child to adult-friend role relationship). (1) In interactions with his teachers at school, Bob was very aware of the importance of doing well and getting good grades; he generally did not initiate interactions, and focused on presenting himself as a good pupil. In language use, he did not seem to take many risks. Even with T1, an extremely supportive teacher who gave very little negative feedback on his language, Bob took few risks. In the following conversation, the study had been in progress for about a year and a half. (Bob's contributions are underlined) 5.33) 59; 30-11-87: 8185-8227 T1: Bob what can I do for you? Bob: I want to publish a book. T1: You want to publish a book. OK. Bob: This one. Ay. I'm now put in there. T1: Now what did you do about your story, my friend? Bob: Zoo. T1: Do you want me to write it or you write it? Bob: You write it. T1: You want me to write it. Now which cover would you like? Blue one or orange one? Which colour? Bob: Blue one. T1: What size do you want it? Do you want it this big or that big? Bob: T1: Bob: T1: Bob: T1: Bob: T1: Bob: This big. This big. OK. Right. So we have that size. Now do you want to write it or do you want me to write it? You write it. Me to write it. OK. You want to put some pictures in it too? Yeah. So we need some white paper and some paper for pictures. Have you got your stories? Yeah. About zoo. What? Zoo.

8
In this conversation, as in the others with his teachers, Bob lets the teacher take the conversational initiative, and limits his responses to utterances with relatively simple syntax. Liu (1991) provides a statistical summary which documents that Bob's interactions with his teachers typically encode a very small range of functions, with sentence structure which is simple relative to that in the other interactional contexts. In interactions with his teachers, Bob rarely initiates, preferring to respond. This pattern is very consistent with Bob's cultural background, in which children are taught to show respect for their teachers, in part by "only speaking when spoken to," and to learn by listening to the teacher. (2) In interacting with peers in the classroom at desk work, Bob, as a highly competitive boy, takes a much more assertive role, initiating interactions, criticizing his classmates' work, demonstrating the superiority of his own, and speaking more fluently. This interactional context is conducive to giving commands, arguing and insulting -- functions which are not culturally appropriate in interactions with the teacher. In the following conversation, Bob and his three friends (Ben, Shayne and Simon) are challenged on the quality of their drawing by a rival group (which includes Mark and Paul); this conversation takes place in July of 1987, six months prior to the conversation with the teacher recorded above: Session 43; July 20, 1987: 5935-5976 Mark: You can't draw the same sorts as in everything you know. Like the original. (Pointing at Ben's work of wave lines) Bob: Yeah yeah. You can draw. Look that. (To Ben.) Mark: You can't can you? So I can't either. Bob: Look he doesn't know how to draw. (To Ben about Mark) Mark: I can draw better than both of you. Ben: Um can you draw peak? (Meaning mountain peak) Mark: Yes I can. Bob: OK you draw. Ben: Can you draw school? Mark: Yes I can. Ben: Can you draw the whole world? Mark: Yes I can. Ben: Ah ha no one can draw the whole world. (Paul comes to the table) Ben: Can you draw the whole world? Mark: He would. Ben: You don't know because you don't know how to draw. Mark: I don't know how to draw the whole world. Bob: You don't know. Ben: Do you know how to draw stars? Bob: No he don't know. Mark: I don't know how to draw stars. There's no one in the world knows how to draw stars. Bob: I know how to draw stars. It's it's very easy. You doesn't know.

9
(Some time later) Paul: Ben I draw better than you. Bob: No you not. It's not. It's teacher's draw. (To Paul) Look look he can't. (To Ben about Paul) Ben: The teacher drawed that. Bob: Yeah. Ben: Bullshit. Bob: Bullshit. Clearly, there are powerful social forces at work in Bob-peer desk work that draw Bob into the conversation; he has established a dominant role in his group from time to time, and he takes the initiative to criticize the work of rivals and to assert his and his friends' superiority. The transcript shows a great deal of linguistic modeling being provided by his peers, and evidence that Bob both notices and incorporates those models into his own utterances in the heated exchanges which are recorded. (See for example line 22-25 above, with the modeling and repetition of "He/you don't know..." and lines 27-28: "x knows how to draw stars")2 Liu's quantitative analyses show a general pattern of greatly increased initiation in Bob-peer interactions, and a much wider range of "functions"3 expressed by Bob in interactions with his peers than with his teachers. (3) The researcher was a friend of the family, and (perhaps more importantly) an adult who (as can be seen in the interactions transcribed in the dissertation) knew how to play and was interested in interacting with Bob and drawing him out in a relaxed home environment. Bob's relationship with the researcher changed from a formal teacher-pupil relationship during the preschool period, to a more relaxed "adult friend" relationship, in which Bob felt free to disobey and even (once or twice) to insult the researcher. The two frequently drew pictures together and played with blocks, talking about their activities as well as about school. Bob tried hard to initiate interactions with the researcher. Bob's attempts to produce English utterances met with the researcher's approval and encouragement (though, interestingly, Liu reports that he was often more judgmental than the teachers, giving Bob more negative feedback on his

An anonymous reviewer points out that it would be interesting to find out how Bob's variable control of AUX 'do' relates to the 'models' he is offered in this and other contexts. 3 Liu defines functions using Nicholas' (1987) categories; see Liu (1991, under review) for discussion.
2

10
language than they did).4 In this third interactional context, Bob used the most multiple constituent and complex structures, and the widest variety of language functions. The following conversation with the researcher took place at roughly the same time as the conversation with the teacher recorded above: Session 61: 10-Dec-87: Bob: Now I know what I can do. Res: Yeah. Bob: I'm not tell you what I'm doing now. Res: Tell me what you're doing. Bob: I'm not tell you. I draw all of them, so it's complete. Res: What do you want to draw? Bob: I'm not tell you. Res: I know. It's nothing except colours. Bob: No. I'm going to draw something. No. I'm not lie to you. Res: I didn't say you were lying. Bob: I'm just draw a ghost. Can't draw. Res: You're drawing a ghost? Bob: Yeah. I can draw all of them. Res: OK. I'll see what sort of ghost you can draw. Bob: I draw a lovely ghost. In the conversational segment transcribed above there appears to be more initiation on Bob's part than in the previous one, with less immediate repetition of modeling. The embedded sentences represented in Now I know what I can do. and I'm not tell you what I'm doing now seem to characterize a learner at a much higher level of proficiency than in the previous segments. The quantitative analysis provided below (and particularly in Liu, 1991) shows that there was a general pattern of increased syntactic complexity, increased initiation, and increased variety of both structure and function in the conversations between Bob and the researcher. The three interactional contexts just described were engaged in during the same period of time, and so during that period, one might suppose that Bob had the same knowledge system of English at his disposal. However, in interaction with the teachers, he only used very simple English, which was a very small part of his knowledge system. In interacting with peers, he made more extensive use of his knowledge by producing more complex English. In interaction with the researcher, he made still more extensive use of this knowledge to the extent of attempting syntactic structures beyond his existing control.

The same reviewer as in Fn. 1 suggests that there may be interesting parallels here between the role played by the researcher and the maximally facilitative adult behavior documented in first language acquisition studies like the Bristol study (Wells 1981).
4

11
Thus, from Liu's (1991) perspective, Bob performs his competence differentially in the three interactional contexts, contexts defined in terms of the differing role relationships the learner had with his interlocutors. A Quantitative Analysis of Bob's IL Performance A number of different quantitative analyses are provided of general ways in which Bob's interlanguage production differed in the three different interactional contexts. Over the same period of time, Bob clearly produced much less language in interactions with his teachers than in the other two contexts: he produced 260 utterances in interactions with his teachers, 2449 in interactions with his peers, and 5716 in interactions with the researcher. This is not surprising, since Bob was only one among many students in the class, and had fewer opportunities to interact with his teachers. What is far more interesting is what the quantitative analyses tell us about the quality of Bob's participation in those three interactional contexts, a quality explainable by the different role relationships involved. For just one of many examples provided in Liu (1991), Bob produced complex structures5 differentially in the context of Bob-peer and Bob-researcher interactions: Table One: Complex Structures Produced by Bob in Two Contexts Context With peers With researcher Total 177 207 (Liu, 1991:211, Table 8.5) Although Bob produces roughly twice as many utterances with the researcher as he does with his peers, he produces 4 times as many complex structures with the researcher as he does with his peers. Other measures show that Bob uses simpler utterances with his teachers than in the other interactions. Bob appeared to be unwilling to take risks with teachers; he seemed to be striving to be correct, and so didn't risk using utterances of which he was unsure. Number 40 82% Percentage 18%

"Complex structures" are just one category of sentence structure catalogued by Liu, and are defined as multiple constituent structures which convey multiple propositions.
5

12
Similarly, Bob initiates and responds to the initiations of others to different degrees in the three contexts:

Table Two: A comparison of initiations and responses in three contexts (between 3/3/87 and 10/12/88) Context With Teachers With Peers W/ Researcher Initiations Responses Total 74 (29%) 186 (71%) 260 (100%) 1798 (73%) 651 (27%) 2449 (100%) 3497 (61%) 2219 (39%) 5716 (100%) (Table 1, Liu, under review)

We have already seen in the excerpted transcripts we have examined that Bob appears to use his interlanguage for different purposes in the context of the three different role relationships. Specifically, he uses his language for a much wider range of speech acts in his interactions with his peers than with his teachers: acts which include criticisms, commands, arguments, insults and even curses. Indeed, many of these speech acts would probably be inappropriate for a Chinese or Australian child to use with an adult who is in an institutional role. Function and form are, of course, inextricably tied. Research on languages for special purposes has clearly shown that when language is used within a given situation for a specific purpose, then particular language forms are called upon to serve that purpose. For different purposes, different language forms or form/function relationships are required. We must expect that when Bob uses his interlanguage to a large extent to (for example) make negative comments about the abilities of his interlocutors, he will have a greater need to use negative and comparative forms (e.g. You can't do X. X is better than you. etc.) in conversations with his peers than in conversations with his teacher. In sum, Liu shows that Bob's use of his interlanguage varies in its general shape as he moves from one situation to another: his use of his interlanguage knowledge is affected by the different interactional contexts, as defined by the different role relationships which pertain. Bob's performance of his competence is different in these different interactional contexts. But such a finding is not new; many other studies have shown that social context, identity of interlocutor, role and task have an impact on learner performance and the language forms which are used for

13
different purposes in different social contexts. In particular, Selinker and Douglas (1985) have made the case that interlanguage performance differs in different discourse domains; learners have been shown to be more fluent and more grammatical when conversing on (for example) their field of study than on other topics. Other researchers (e.g. Zuengler 1989, Whyte in press) have found the discourse domain model to be very persuasive. The crucial question for us now, is this: does Liu's data show that this variation in interlanguage performance has any impact on, or relationship to, the learner's acquisition, over time, of any specific structure of the interlanguage? Effect of Interactional Context on Acquisition Liu's study goes beyond use and performance to make claims about the way in which interactions in the three contexts studied differentially affected acquisition. Liu claims that the interactions in which Bob engaged had an impact on his interlanguage development, not only in terms of the rate of that development, but also on the route of his interlanguage development. Let us examine each of these claims in turn. Definitions It has already been observed that Liu defines interactional context in terms of the interlocutors with whom Bob interacts. The crucial thing here is the role relationships which obtain with those interlocutors. A role is a very useful construct for our purposes, in that it is simultaneously social and cognitive: it is defined both by the learner's perception of his relationship to others and by social norms which the learner has internalized. It is thus a very helpful construct for SLA researchers to use in examining the relationship between the learner's external social context and his internal cognitive processes. Liu, following Pienemann & Johnston (1987), defines Bob's acquisition of a form as occurring at the point of first use of that form. Different researchers define "point of acquisition" differently; which definition is "correct" is an unresolved question in our field and we will not resolve it here. What is important for our purposes is to simply note that Liu, following Pienemann & Johnston (1987), opts for first use of a form as marking the acquisition of that form. Rate.

14
Liu (1991) uses Pienemann and Johnston's (1987) framework for the development of English as a second language to examine the development of various aspects of Bob's interlanguage grammar. He focuses for example upon the developmental stages in Bob's use of interrogative forms (in the sequence which follows, the examples are from Bob's data and are marked by the session in which they occur):

Stage l: Single word Stage 2: SVO?

You like number one? (Session 29) Stage 3: Do-front without inversion WHX-front without inversion Why you do that? (Session 36) Stage 4: Pseudo-inversion Y/N inversion Where's the monkey? (Session 24) Stage 5: Aux-2nd Suppletion What are you doing? (Session 24) Stage 6: Question tag You don't like green, are you? (Session 49) Liu's data show a clear and consistent pattern whereby almost every new stage in this sequence (all except Stage 3, about which more below) appears first in interaction with the researcher, and only much later in other settings. For example, Stage 5 interrogative forms initially appear in Bob-researcher interactions in Session 23, but do not appear in Bob-peer interactions until Session 36, and last in Bobteacher interactions. Stage 6 interrogative forms appear first in Bob-researcher interactions in Session 49, but do not appear in Bob-peer interactions until Session 76, nearly nine months later. In general, new structures appear first in the interactions between Bob and the researcher, spread to the interactions with his peers, and appear last in his interactions with his teacher. Liu suggests that the context in which a form is first used is the context in which that form was acquired. His analysis suggests that different interactional contexts support different rates of development of particular interlanguage features. In other words, if the learner is deprived of the opportunity to interact in certain contexts and role relationships, then his rate of IL development will be slower. Liu states that if Bob had not had the

15
opportunity to interact with the researcher, the overall rate of development of his interlanguage is, in fact, likely to have been much slower. Why Some Interactional Contexts Promote Faster Acquisition Liu argues that Bob pushes the limits of his competence in interaction with the researcher, thus, we might suggest, making his competence more "permeable" (Adjemian 1976) and open to change in that context. Bob takes risks with his interlanguage production differentially in the three different interactional contexts. In interacting with the researcher, as opposed to other interlocutors, Bob speaks in a wider range of styles ranging from narratives, discussions to arguments. Bob also produces a wider variety of forms in interaction with the researcher than in the other interactional contexts.6 Further, in speaking with this interested and supportive adult friend, Bob's interlanguage was characterized by inaccurate syntax resulting from his exploitation of his competence in English to its limits. The consequence was that he produced much more complex structures laden with non-target forms, in this context than in any other context. (Liu, 1991:227) For example, at the time Session 35 was taped, Bob was consistently responding to teacher questions in the classroom with "I don't know", but in Session 35 with the researcher he produced responses to questions such as the following:

Session 35; 28-May-87: (Bob and the researcher are looking at work done by pupils) Researcher: Where's yours? Bob: NI ZHI DAO NA SHI HOU MA? (You know that time?) NA SHI HOU WO HAI MEI LAI NE. (It was the time when I hadn't started school yet.) Longer longer. I'm not come. Go to go to school. It's. Is mine.
6

not

For example, he produces a wider range of Stage 5 interrogative forms with the researcher than with other interlocutors: What do you need? What color do you like? What do you want to do? What are you doing? Where did I put it? Which do you want to color in? How do you know? How do people go up the top? How can we do it? Who's got long?

16
Researcher: Longer longer? Bob: SHEN ME? (What?) Longer longer. I can't read. I can't draw picture cause come. Here Bob initially tries to speak in Chinese to the researcher, who does not respond in Chinese. Then Bob resorts to communication strategies (cf. Tarone 1978; Faerch & Kasper 1983) in his attempts to specify a time frame in English (something like "long ago, when I had not started school yet"). He creates some interlanguage vocabulary items, such as "longer longer" (word coinage) for "a long time ago", and constructions such as "I'm not come" (an approximation) for "I hadn't gone to school yet." These forms are not related to Chinese sentence constructions, so are not based on language transfer. They are not targetlike, either. Their creation and use indicates a willingness to take extended turns in interaction and to stretch the interlanguage system beyond "safe" limits -- in short, to take risks in creating forms outside the interlanguage system in attempting to communicate. What do Liu's data tell us about the way in which interactional dynamics may affect internal L2 acquisition processes? Another way to ask this question is to ask: what are the forces that make some learners' interlanguage stay permeable, keep developing; what prevents fossilization? Clearly, there are many cases in which innate, universal cognitive processes alone do not just go into operation in response to input and develop inexorably to the finish. Swain (1985) has shown that the ILs of immersion children, exposed to plentiful L2 input data from teachers, nevertheless stop developing. She has argued that in situations (like immersion classrooms) where learners are not required to produce L2, their interlanguages fossilize, cease developing. Bob received English input from teachers and peers, yet those were not the contexts in which his interlanguage seemed to be most stimulated to move from one stage to another. Liu's data show us first that interactional demands differ in different social contexts, and argues further that ONE context can be identified as the one in which the learner's interlanguage system is most unstable, most variable, and most open to change -- and in which the first elements of more advanced developmental stages are produced. There is thus evidence that this is the context in which acquisition is most likely to have occurred.

I'm not

17
What exactly is it that might make this context so positive for IL development? The answer to this question must at present require some speculation as to the possible cognitive effects of this interactive context. There are at least two possibilities which we may be able to explore in a closer examination of data such as these. It is possible that the researcher exemplifies the ideal interlocutor, one who provides input which is most suited to the learner's developmental needs. The researcher clearly is the interlocutor who focuses solely upon the learner and communicating with him. Thus, it is possible that the researcher provides the input which is most finely attuned to Bob's needs, and that this explains why researcher-Bob interactions are so productive. Another possibility, alluded to above, is that the key variable is Bob's attempt to produce comprehensible output (Swain 1985). Bob is given much more encouragement and opportunity to produce the L2 and to elaborate his utterances in the Bob-researcher context than in the Bob-peer context, where Bob has to compete for the floor with several peers. In this view, it is in those interactional contexts where the learner needs to produce output that the current interlanguage system cannot handle that the learner pushes the limits of that interlanguage system to MAKE it handle that output, thus keeping that system permeable and open to change. In such contexts, the learner functions in much the same way as the learner in Schmidt & Frota (1985): struggling to produce output, becoming aware of a gap or need for a structure, and THEN noticing that structure in the input. Liu's data can be argued to show that it is precisely in those contexts where Bob has to produce output that his IL cannot handle that the IL develops fastest, with the richest variety of IL structures and even possibly with structures out of developmental sequence (see Route below). This explanation also allows us to see how communication strategies and interlanguage variation may be related, both to one another and to the acquisition process. In this view, a learner's use of communication strategies can function to compensate for the current limits of an IL, resulting in free variation as the learner tests new hypotheses in the search for an appropriate word or structure. A learner in search of the right word or structure is a learner who is open to noticing such things in the input (Schmidt & Frota, 1985).

18
Liu's data are convincing evidence that an exploration of the causes of interlanguage variation can shed light upon the process of second-language acquisition. The interactions in which Bob engages have an impact upon the interlanguage forms he is willing to use at any given time, and the degree to which he is willing to go beyond the limits of his interlanguage competence. This willingness in turn may have an impact upon the rate of his interlanguage development. His interlanguage develops differentially in different interactional contexts, as he enters into different role relationships and responds to different demands from his interlocutors. Route We have argued that certain types of interaction encourage faster and more complete development of certain features in the interlanguage than do others. Of course, Selinker and Douglas, 1985, and others have argued this same point. But Liu's is the first study of which we are aware which examines the development of a specific grammatical form over time in several contexts and provides specific data in support of this argument. Liu uses the development of interrogative forms in Bob's interlanguage to show that interactional context can affect, not just the rate of acquisition, but particular sequences of development in the interlanguage. Pienemann and Johnston (1987) claim that there is a universal sequence of acquisition of interrogative forms, such that Stage 4 or 5 features should not appear before Stage 3 features. In Pienemann & Johnston's theory, a learner cannot progress to a higher level stage of acquisition without first going through ALL the lower level stages of acquisition of any given structure. The stages proposed in this theory are claimed to be universal. Although there may be some disagreement on some of the details of Pienemann and Johnston's theory (cf. Hudson 1993), in fact there is general agreement among SLA researchers on the broad outlines of this sequence of acquisition of questions in English, as noted by Larsen-Freeman & Long (1991:93): it is generally agreed that questions with non-inversion of the SVO pattern (e.g. P&J's Stage 3) occur BEFORE questions in which inversion occurs (as in P&J's Stages 4 and 5). Yet in Bob's case, Stage 4 and 5 interrogative forms did emerge in interactions with the researcher long before Stage 3 forms emerged.

19
The data show that Stages 4 and 5 interrogative forms emerged in Bob's interaction with the researcher in Sessions 23 and 24. Stage 3 interrogative forms did not emerge until nearly three months later and were first used in Session 36 in Bob's interaction with his peers at school. Stage 3 interrogative forms started to appear in Bob's interaction with the researcher in Session 40. It would appear, then, that Bob's participation in the interactional context with the researcher altered a so-called universal sequence of acquisition of interrogatives in Bob's IL. Liu (1991) suggests that it is the researcher's intensive use of Stage 4 and 5 forms in the input, together with an increased opportunity for Bob to produce such forms, which influenced Bob to acquire those forms first in that context. Bob's interaction with the researcher gave him a context where there were opportunities and incentive for him to acquire and use those Stage 4 and 5 forms before Stage 3 forms. This claim implies that external social demands can be so strong that they can cause an alteration in internal psychologically-motivated sequences of acquisition -- even so-called universal sequences of acquisition. Liu's (1991) analysis of Bob's sequence of acquisition is based upon Pienemann and Johnston's criterion of first use of a form as marking that form's acquisition. Clearly, we need to examine Bob's variable use of question forms in more detail, using tools such as VARBRUL, in order to see how frequent these forms are in the three interactional contexts over time. In addition, we need to examine the development of other structures in Bob's interlanguage, such as negation and relative clause formation, in order to see whether interactional context affects the development of those structures in a similar way. However, Liu's research is important in that it points a direction for future research. The study of Bob's language development illustrates the kind of data we need to obtain in an attempt to determine the extent to which interactional context may be able to override or alter any claimed "innate universal sequence of acquisition". Such an investigation could shed light upon the delicate relationship which undoubtedly exists in SLA between internal "innate" and external "contextual" forces. And clearly, any adequate theory of second-language acquisition must be able to account for that relationship between innate and contextual forces which occurs in the acquisition process.

20

References Adjemian, C. (1976) On the nature of interlanguage systems. Language Learning 26:297-320. Beebe, L. (1980). Sociolinguistic variation and style-shifting in second language acquisition. Language Learning 30,2:433-447. Dickerson, L. (1975). The learner's interlanguage as a system of variable rules. TESOL Quarterly 9:401-407. Eckman, F. (1991) The competence-performance issue in second-language acquisition theory: A debate. Paper presented at the Michigan State University Conference on Theory Construction and Methodology in SLA, East Lansing Michigan. Forthcoming in E. Tarone, S. Gass & A. Cohen (Eds.) Research Methodology in Second Language Acquisition. Erlbaum Publishers. Faerch, C. & G. Kasper (Eds.). (1983) Strategies in Interlanguage Communication. London: Longman. Flynn, S. & W. O'Neil (Eds.) (1990). Linguistic Theory in Second Language Acquisition. Dordrecht: Reidel Press. Gass, S. & J. Schachter (Eds.) (1989). Linguistic Perspectives on Second Language Acquisition. New York: Cambridge University Press. Gass, S. (1988) Integrating research areas: A framework for secondlanguage studies. Second Language Research 9, 2: 99-117. Gregg, K. (1990). The variable competence model of second language acquisition and why it isn't. Applied Linguistics 11,4:364-383. Gregg, K. (1993). Taking explanation seriously; or, Let a couple of flowers bloom. Applied Linguistics 14/3:276-294. Hatch, E. (1978) Discourse analysis and second language acquisition, in E. Hatch (Ed.) Second Language Acquisition: A Book of Readings. (pp. 402435) Rowley, Mass: Newbury House.

21
Hudson, T. (1993) Nothing does not equal zero: Problems with applying developmental sequence findings to assessment and pedagogy. Studies in Second Language Acquisition 15,4:461-494. Larsen-Freeman, D. & M. Long (1991). An Introduction to Second Language Acquisition Research. London: Longman. Liu, G. (under review). Interaction and SLA: A case study. Studies in Second Language Acquisition. Liu, G. (1991). Interaction and second language acquisition: A case study of a Chinese child's acquisition of English as a Second Language. Unpublished Ph.D. Thesis, La Trobe University. Long, M. (1985) Input and second language acquisition theory. In Gass & Madden (eds.) Input in Second Language Acquisition (377-393) Rowley, Mass: Newbury House. Markee, N. (1991) Toward an ethnomethodological respecification of second-language acquisition studies. Paper presented at the Michigan State University Conference on Theory Construction and Methodology in SLA, East Lansing, Michigan. Forthcoming in Tarone, Gass & Cohen (Eds.), Research Methodology in Second Language Acquisition. Erlbaum Publishers. Nicholas, H. (1987) A comparative study of the acquisition of German as a first and a second language. Unpublished Ph. D Thesis, Monash University. Pica, T. & C. Doughty (1985). Input and interaction in the communicative language classroom: a comparison of teacher-fronted and group activities. In Gass & Madden (eds.), Input in Second Language Acquisition (115-132). Rowley, Mass: Newbury House. Pienemann, M. & M. Johnston (1987). Factors influencing the development of language proficiency. In D. Nunan (ed.), Applying Second Language Acquisition Research (45-141). Adelaide, Australia: National Curriculum Resource Centre. Preston, D. (1989) Sociolinguistics and Second Language Acquisition. Oxford: Blackwell. Rutherford, W. (1984). Language Universals and Second Language Acquisition. Amsterdam: John Benjamins. Sato, C. (1990) The Syntax of Conversation in Second Language Development. Tubingen: Gunter Narr. Schmidt, R. & S. Frota (1986). Developing basic conversational ability in a second language: a case study of an adult learner of Portuguese. In Day, R. (ed.) Talking to Learn (pp. 237-326). Rowley, Mass: Newbury House. Schmidt, R. (1994). Paper presented in Colloquium on Awareness in SLA at the American Association for Applied Linguistics Conference, Baltimore, Maryland, February 1994. Schumann, J. (1978). The Pidginization Process: A Model for Second Language Acquisition. Rowley, Mass: Newbury House. Selinker, L. & D. Douglas (1985) Wresting with 'context' in interlanguage theory. Applied Linguistics 6:190-204. Swain, M. (1985) Communicative competence: some roles of comprehensible input and output in its development. In S. Gass & C. Madden (Eds.) Input in Second Language Acquisition. (pp. 235-253) Newbury House. Swain, M. (in press) The output hypothesis: Just speaking and writing aren't enough. Canadian Modern Language Review 50,1. Tarone, E. (1978) Conscious communication strategies in interlanguage: a progress report. In Brown, H.D., C. Yorio & R. Crymes (Eds.) On TESOL '77 (194-203). Washington D.C.: TESOL. Tarone, E. (1979) Interlanguage as chameleon. Language Learning 29,1:181-191. Tarone, E. (1983) On the variability of interlanguage systems. Applied Linguistics 4,2:143-163. Tarone, E. (1988). Variation in Interlanguage. London: Edward Arnold. Tarone, E. (1990) On variation in interlanguage: A response to Gregg. Applied Linguistics 11,4:392-400. Wells, G. (1981). Learning through Interaction: The Study of Language Development. Cambridge: Cambridge University Press. White, L. (1989). Universal Grammar and Second Language Acquisition. Amsterdam: John Benjamins. Whyte, S. (in press) Acquisition in context: The effect of topic knowledge on second language. In L. Bouton & Y. Kachru (Eds.), Pragmatics and Language Learning (Vol. 5). Urbana, IL: DEIL, University of Illinois. Wong Fillmore (1976) The second time around: Cognitive and social strategies in second language acquisition. Unpublished Ph.D. dissertation, Stanford University.

22
Young, R. (1991) Variation in interlanguage morphology. New York: Lang. Zuengler, J. (1989) Performance variation in NS-NNS interactions: Ethnolinguistic difference or discourse domain? In S. Gass, C. Madden, D. Preston, & L. Selinker (Eds.) Variation in second language acquisition: Discourse and pragmatics (pp. 228-244). Clevedon, UK: Multilingual Matters.