In all probability, one of the main reasons for low achievement by many classroom SL learners is simply that they do not have enough time to Practice the new language. This is especially serious in large EFL classes in which students need to develop aural-oral skills, but it is also relevant to the ESL context.

From observational studies of classrooms (e.g., Hoetker and Ahlbrand 1969 and Fanselow 1977), we know that the predominant mode of instruction is what might be termed the lockstep, in which one person (the teacher) sets the same instructional pace and content for everyone, by lecturing, explaining a grammar point, leading drill work, or asking questions of the whole class. The same studies show that when lessons are organized in this manner, a typical teacher of any subject talks for at least half, and often for as much as two thirds, of any class period (Flanders 1970). In a 50-minute lesson, that would leave 25 minutes for the students. However, since 5 minutes is usually spent on administrative matters (getting pupils in and out of the room, calling the roll, collecting and distributing homework assignments, and so on) and (say) 5 minutes on reading and writing, the total time available to students is actually more like 15 minutes. In an EFL class of 30 students in a public secondary school classroom, this averages out to 30 seconds per student per lesson-or just one hour per student per year. An adult ESL student taking an intensive course in the United States does not fare much better. In a class of 15 students meeting three hours a day, each student will have a total of only about one and a half hours of individual practice during a six-week program. Contrary to what some private language school advertisements would have us believe, this is simply not enough.

An unfortunate but hardly surprising side effect of this sort of pseudo- communication is that students' attention tends to wander. Consequently, teachers maintain a brisk pace to their questions and try to ensure prompt and brief answers in return. This is usually quite feasible, since what the students say requires little thought (the same question often being asked several times) and little language (mostly single phrases or short "sentences"). Teachers quickly "correct" any errors, and students appreciate just as quickly that what they say is less important than how they say it.

Such work may be useful for developing grammatical accuracy (although this has never been shown). It is unlikely, however, to promote the kind of conversational skills students need outside the classroom, where accuracy is often important but where communicative ability is always at a premium.

Group work can help a great deal here. First, unlike the lockstep, with its single, distant initiator of talk (the teacher) and its group interlocutor (the students), face-to-face communication in a small group is a natural setting for conversation. Second, two or three students working together for five minutes at a stretch are not limited to producing hurried, isolated "sentences." Rather, they can engage in cohesive and coherent sequences of utterances, thereby developing discourse competence, not just (at best) a sentence grammar. Third, as shown by Long, Adams, McLean, and Castanos (1976), students can take on roles and adopt positions which in lockstep work are usually the teacher's exclusive preserve and can thus practice a range of language functions associated with those roles and positions. While solving a problem concerning the siting of a new school in an imaginary town, for example, they can suggest, infer, qualify, hypothesize, generalize, or disagree. In terms of another dimension of conversational management, they can develop such skills-also normally practiced only by the teacher--as topic-nomination, turn-allocation, focusing, summarizing, and clarifying. (Some of these last skills also turn out to have considerable psycholinguistic importance.) Finally, given appropriate materials to work with and problems to solve, students can engage in the kind of information exchange characteristic of communication outside classrooms--with all the creative language use and spontaneity this entails-wbere the focus is on meaning as well as form. In other words, they can in all these ways develop at least some of the variety of skills which make up communicative competence in a second language.

In addition to this kind of variability in specific SL abilities, other kinds of individual differences ignored by lockstep teaching include students' age, cognitive/ developmental stage, sex, attitude, motivation, aptitude, personality, interests, cognitive style, cultural background, native language, prior language learning experience, and target language needs. In an ideal world, these differences would all be reflected, among other ways, in the pacing of instruction, in its linguistic and cultural content, in the level of intellectual challenge it poses, in the manner of its presentation (e.g., inductive or deductive), and in the kinds of classroom roles students are assigned.

Group work obviously cannot handle all these differences, for some of which we still lack easily administered, reliable measures. Once again, however, it can help. Small groups of students can work on different sets of materials suited to their needs. Moreover, they can do so simultaneously, thereby avoiding the risk of boring other students who do not have the same problem, perhaps because they speak a different first language, or who do have the same problem but need less time to solve it. Group work, then, is a first step toward individualization of instruction, which everyone agrees is a good idea but which few teachers or textbooks seem to do much about.

Many students, especially the shy or linguistically insecure, experience considerable stress when called upon in the public arena of the lockstep classroom. This stress is increased by the knowledge that they must respond accurately and above all quickly. Research (see, for example, Rowe 1974 and White and Lightbown 1983) has shown that if students pause longer than about one second before beginning to respond or while making a response, or (worse) appear not to know the answer, or make an error, teachers will tend to interrupt, repeat, or rephrase the question, ask a different one, "correct," and/or switch to another student. Not all teachers do these things, of course, but most teachers do so more than they realize or would want to admit.

In contrast to the public atmosphere of lockstep instruction, a small group of peers provides a relatively intimate setting and, usually, a more supportive environment in which to try out embryonic SL skills. After extensive research in British primary and secondary school classrooms, Barnes (1973:19) wrote of the smallgroup setting:

An intimate group allows us to be relatively inexplicit and incoherent, to change direction in the middle of a sentence, to be uncertain and selfcontradictory. What we say may not amount to much, but our confidence in our friends allows us to take the first groping steps towards sorting out our thoughts and feelings by putting them into words. I shall call this sort of talk "exploratory."

In his studies of children's talk in small groups, Barnes found a high incidence of pauses, hesitations, stumbling over new words, false starts, changes of direction, and expressions of doubt I think probably, and so on). This was the speech of children "talking to learn" (Barnes 1973:20) -talking, in other words, in a way and for a purpose quite different from those which commonly characterize interaction in a full-class session. There, the "audience effect" of the large class, the perception of the listening teacher as judge, and the need to produce a short, polished product all serve to inhibit this kind of language.

It is not only size and lack of intimacy that discourage exploratory talk: if relationships have been formalized until they approach ritual, this, too, will make it hard for anyone to think aloud. Some classrooms can become like this, especially when the teacher controls very thoroughly everything that is said.

Several advantages have already been claimed for group work. It allows for a greater quantity and richer variety of language practice, practice that is better adapted to individual needs and conducted in a more positive affective climate. Students are individually involved in lessons more often and at a more personal level. For all these reasons and because of the variety group work inevitably introduces into a lesson, it seems reasonable to believe that group work motivates the classroom learner.

Empirical evidence supporting this belief has been provided by several studies reported recently in Littlejohn (1983). It has been found, for example, that small-group, independent study can lead to increased motivation to study Spanish among beginning students (Littlejohn 1982); learners responding to a questionnaire reported that they felt less inhibited and freer to speak and make mistakes in the small group than in the teacher-led class. Similarly, in a study of children's attitudes to the study of French in an urban British comprehensive school (Fitz-Gibbon and Reay 1982), three quarters of the pupils ranked their liking for French as a school subject significantly higher after completing a program in which 14-year old non-native speakers tutored 11-year-old non-natives in the language.

It has further been shown that NSs, especially those (like ESL teachers) with considerable experience in talking to foreigners, are adept at modifying not just the language itself, but also the shape of the conversations with NNSs in which the modified speech occurs. They help their non- native conversational partners both to participate and comprehend in a variety of ways. For example, they manage to make topics salient by moving them to the front of an utterance, saying something like San Diego, did you like it?, rather than Did you like San Diego? They use more questions than they would with other NSs and employ a number of devices for clarifying both what they are saying and what the NNS is saying. The devices include clarification requests, confirmation checks, comprehension checks, and repetitions and rephrasings of their own and the N NSs' utterances. (For a review of the research on conversa.tional adjustments to NNSs, see Long 1983a.)

It is important to note that when making these linguistic and conversational adjustments, speakers are concentrating on communicating with the NNS; that is, their focus is on what they are saying, not on how they are saying it. As with parents and elder siblings talking to young children, the adjustments come naturally from trying to communicate. While their use seems to grow more sophisticated with practice, they require no special training.

A recent study by Hawkins (in press) has shown that it is dangerous to assume that the adjustments always lead to comprehension by NNSs, even when they appear to have understood, as judged by the appropriateness of their responses. On the other band, at least two studies (Chaudron 1983 and Long in press) have demonstrated clear improvements in comprehension among groups of NNSs as a result of specific and global speech modifications, respectively. Other research has demonstrated that the modifications themselves are more likely to occur when the native speaker and the non-native speaker each start out a conversation with information the other needs in order for the pair to complete some task successfully. Tasks of this kind, called two-way tasks (as distinct from one-way tasks, in which only one speaker has information to communicate), result in significantly more conversational modifications by the NS (Long 1980, 1981, 1983b). This is probably because the need for the NS to obtain unknown information from the NNS makes it important for the NS to monitor the NNS's level of comprehension and thus to adjust until the NNS's understanding is sufficient for performance of his or her part of the task.

There is also a substantial amount of evidence consistent with the idea that the more language that learners hear and understand or the more comprehensible input they receive, the faster and better they learn. (For a review of this evidence, see Krashen 1980, 1982 and Long 1981, 1983b.) Krashen has proposed an explanation for this, which he calls the Input Hypothesis, claiming that learners improve in a SL by understanding language which contains some target language forms (phonological, lexical, morphological, or syntactic) which are a little ahead of their current knowledge and which they could not understand in isolation. Ignorance of the new forms is compensated for by hearing them used in a situation and embedded in other language that they do understand:

A necessary condition to move from stage i to stage i + 1 is that the acquirer understand input that contains i + 1, where "understand" means that the acquirer is focused on the meaning and not the form of the utterance (Krasben 1980:170).

Whether or not simply hearing and understanding the new items are both necessary and sufficient for a learner to use them successfully later is still unclear. Krashen claims that speaking is unnecessary, that it is useful only as a means of obtaining comprehensible input. However, at least one researcher (Swain in press) has argued that learners must also be given an opportunity to produce the new forms--a position Swain calls the "comprehensible output [italics added] hypothesis." What many researchers do agree upon is that learners must be put in a position of being able to negotiate the new input, thereby ensuring that the language in which it is heard is modified to exactly the level of comprehensibility they can manage.

As noted earlier, the research shows that this kind of negotiation is perfectly possible, given two- way tasks, in NS/NNS dyads. The problem for classroom teachers, of course, is that it is impossible for them to provide enough of such individualized NS/NNS opportunities for all their students. It therefore becomes essential to know whether two (or more) non-native speakers working together during group work can perform the same kind of negotiation for meaning. This question has been one of the main motivations for several recent studies of NNS/NNS conversation, often referred to in the literature as interlanguage talk. The focus in these studies of NNSs working together in small groups is no longer just the quantity of language practice students are able to engage in, but the quality of the talk they produce in terms of the negotiation process.

An early study of interlanguage talk was carried out by Long, Adams, McLean, and Castahos (1976) in intermediate-level, adult ESL classes in Mexico. The researchers compared speech samples from two teacher-led class discussions to speech from two smallgroup discussions (two learners per group) doing the same task. To examine the quantity and quality of speech in both contexts, the researchers first coded moves according to a special category system designed for the study. Quality of speech was defined by the variety of moves, and quantity of speech was defined by the number of moves. The amount and variety of student talk were found to be significantly greater in the small groups than in the teacher-led discussions. In other words, students not only talked more, but also used a wider range of speech acts in the small-group context.

In a larger study, Porter (1983) examined the language produced by adult learners in task- centered discussions done in pairs. The learners were all N.Ss of Spanish. The 18 subjects (12 NNSs and 6 NSs) represented three proficiency levels: intermediate, advanced, and native speaker. Each subject participated in separate discussions with a subject from each of the three levels. Porter was thus able to compare interlanguage talk with talk in NS/NNS conversations, as well as to look for differences across learner proficiency levels. Among many other findings, the following are relevant to the present discussion:

1. With regard to quantity of speech, Porter's results supported those of Long, Adams, McLean, and Castaiios (1976): Learners produced more talk with other learners than with NS partners. In addition, learners produced more talk with advanced-level than with intermediate-level partners, in part because the conversations with advanced learners lasted longer.

3. Other analyses focused on the interactional features of the discussions; no significant differences were found in the amount of repair by NSs and learners. Repair was a composite variable, consisting of confirmation checks, clarification requests, comprehension checks, and three communication strategies (verification of meaning, definition request, and indication of lexical uncertainty). Porter emphasized the importance of this finding, suggesting that it shows that learners are capable of negotiating repair in a manner similar to NSs and that learners at the two proficiency levels in her study were equally competent to do such repair work. A related and not surprising finding was that learners made more repairs of this kind with intermediate than with advanced learners.

4. Closer examination of communication strategies, a subset of repair features, revealed very low frequencies of "appeals for assistance" (Tarone 1981), redefined for the Porter study to include verification of meaning, definition request, and indication of lexical uncertainty. In addition, learners made the appeals in similar numbers whether talking to NSs or to other learners (28 occurrences in four and a quarter hours with NSs versus 21 occurrences in four and a half hours with other learners.) Porter suggested that her data contradict the notion that other NNSs are not good conversational partners because they cannot provide accurate input when it is solicited. In fact, however, learners rarely ask for help, no matter who their interlocutors may be. it would appear that the social constraints that operate to keep foreigner-talk repair to a minimum (McCurdy 1980) operate similarly in NNS/NNS discussions.

5. Further evidence of these social constraints is the low frequency of other- correction by both learners and NSs. Learners corrected 1.5 percent and NSs corrected 8 percent of their interlocutors' grammatical and lexical errors. Also of interest is the finding that learners miscorrected only .3 percent of the errors their partners made, suggesting that miscorrections are not a serious threat in unmonitored group work.

6. The findings on repair were paralleled by those on another interactive feature, labeled prompts, that is, words, phrases, or sentences added in the middle of the other speaker's utterance to continue or complete that utterance. Learners and NSs provided similar numbers of prompts. One significant difference, however, was that learners prompted other learners five times more than they prompted NSs; thus, learners got more practice using this conversational resource with other learners than they did with NSs.

Overall, Porter concluded that although learners cannot provide each other with the accurate grammatical and sociolinguistic input that NSs can, learners can offer each other genuine communicative practice, including the negotiation for meaning that is believed to aid SLA. Confirmation of Porter's findings has since been provided in a small-scale replication study by Wagner (1983).

Two additional studies of interlanguage talk (Varonis and Gass 1983, Gass and Varonis in press) should be mentioned. In the first study, the researchers compared interlanguage talk in 11 non-native conversational dyads with conversation in 4 NS/NNS dyads and 4 NS/NS dyads. Like the learners in Porter's (1983) study, the NNSs were students from two levels of an intensive English program; unlike Porter's subjects, these learners were from two native language backgrounds (Japanese and Spanish). Varonis and Cass tabulated the frequency of what they term nonunderstanding routines, which indicate a lack of comprehension and lead to negotiation for meaning through repair sequences.

The main finding in the Varonis and Gass study was a greater frequency of negotiation sequences in non-native dyads than in dyads involving NSs. The most negotiation occurred when the NNSs were of different language backgrounds and different proficiency levels; the next highest frequency was in pairs sharing a language or proficiency level; and the lowest frequency was in pairs with the same language background and proficiency level. On the basis of these findings, Varonis and Gass argue for the value of non-native conversations as a nonthreatening context in which learners can practice language skills and make input comprehensible through negotiation.

Building on this study, Cass and Varonis (in press) next examined negotiation by NNSs in two additional communication contexts: what Long (1981) calls one-way and two-way tasks. In the one- way task, one member of a dyad or triad described a picture which the other member(s) drew. In the two-way task, each member heard different information about a robbery, and the dyad/triad was to determine the identity of the robber. The participants, who were grouped into three dyads and one triad, were nine intermediate students from four different language backgrounds in an intensive ESL program.

Cass and Varonis looked for differences in the frequency of negotiation sequences across the two task types; they found that there were more indicators of nonunderstanding in the one-way task, but the difference was not statistically significant. They suggest that there may have been more need for negotiation on the one-way task because of the lack of shared background information. A second concern in the study was the role of the participant initiating the negotiation. The finding was not surprising: The student drawing the picture in the one-way task used far more indicators of nonunderstanding than the describer did. A third finding related to the oneway task was a decrease in the number of nonunderstanding indicators on the second trial: Familiarity with the task seemed to decrease the need for negotiation, even though the roles were switched, with the students doing the describing and those doing the drawing changing places.

As in their earlier study, Gass and Varonis argue that negotiation in non-native exchanges is a useful activity in that it allows the learners to manipulate input. When input is negotiated, they naintain, conversation can then proceed with a minimum of confusion; additionally, the input will be more meaningful to the learners because of their involvement in the negotiation process.

The importance of learners' being able to adjust input by providing feedback on its comprehensibility was also stressed by Gaies (1983b). Gaies examined learner feedback to teachers on referential communication tasks. The participants were ESL students of various ages and proficiency levels and their teachers, grouped into 12 different dyads and triads. The students were encouraged to ask for clarification or re-explanation wherever necessary to complete the task of identifying and sequencing six different designs described by the teacher. On the basis of the audiotaped data, Gaies developed an inventory of learner verbal feedback consisting of 4 basic categories (responding, soliciting, reacting, and structuring) and 19 subcategories. Of interest here are Gaies' findings that 1) learners used a variety of kinds of feedback, with reacting moves being the most frequent and structuring moves the least frequent, and 2) learners varied considerably in the amount of feedback they provided.

In another study of non-native talk in small-group work, this time in a classroom setting, Pica and Doughty (in press) compared teacher-fronted discussions and small-group discussions on (oneway) decision-making tasks. Their data were taken from three classroom discussions and three small- group discussions (four students per group) involving low- inter m ediate-level ESL students. Their findings on grammaticality and amount of speech are similar to those of Porter (1983). Pica and Doughty found that student production, as measured by the percentage of grammatical T-units (Hunt 1970) per total number of T-units, was equally grammatical in the two contexts. In other words, students did not pay closer attention to their speech in the teacher's presence. In terms of the amount of speech, Pica and Doughty found that the individual students talked more in their groups than in their teacber-fronted discussions, confirming previous findings of a clear advantage for group work in this area.

Pica and Doughty also examined various interactional features in the discussions. They found a very low frequency of comprehension and confirmation checks and clarification requests in both contexts and pointed out that such interactional negotiation is not necessarily useful input for the entire class, as it is usually directed by and at individual students. In the teacber-led context, it serves only as a form of exposure for other class members, who may or may not be listening, whereas such negotiated input directed at a learner in a small group is far more likely to be useful for that learner. Finally, an examination of other- corrections and completions showed those features to be more typical of group work than of teacher-led discussions, thus supporting the arguments for learners' conversational competence made by Porter and by Varonis and Gass.

In a follow-up study, Doughty and Pica (1984) compared language use in teacher-fronted lessons, group work (four students per group), and pair work on a two-way task. The participants, who had the same level of proficiency as those in Pica and Doughty (in press), bad to give and obtain information about bow flowers were to be planted in a garden. Each started with an individual felt board displaying a different portion of a master plot. At the end of the activity, all participants were supposed to have constructed the same picture, which they compared against the master version then shown to them for the first time. The researchers compared their findings with those from their earlier study, in which a one-way task had been used.

Doughty and Pica also noted that negotiation work as a percentage of total talk was lower in teacher-fronted lessons on both oneway and two-way tasks. This finding, they suggested, may indicate that students are reluctant to indicate a lack of understanding in front of their teacher and an entire class of students and for that reason do not negotiate as much comprehensible input in wholeclass settings. This suggestion was supported by the researchers' informal assessment of students' actual comprehension, as judged by their lower success rate on the garden-planting task in the teacher-led than in the small-group discussions. Doughty and Pica concluded by emphasizing the importance not of group work per se, but of the nature of the task which the teacher provides for work done in small groups.

Finally, two nonquantitative studies have contributed insights into interlanguage talk. Bruton and Samuda (1980) studied errors and error treatment in small-group discussions based on various problemsolving tasks. Their learners were adults from a variety of language backgrounds, studying in an intensive course. The main findings were that 1) learners were capable of correcting each other successfully, even though their teachers had not instructed them to do so, and 2) learners were able to employ a variety of different errortreatment strategies, among which were the offering of straight alternatives (i.e., explicit corrections) and the use of repair questions.

In general, the learners' treatments were much like those of their teachers, except that the most frequent errors treated by the learners were lexical items, not syntax or pronunciation. Bruton and Samuda also noted that in ten hours of observation, only once was a correct item changed to an incorrect one by a peer; furthermore, students did not pick up many errors from each other, a finding also reported by Porter (1983). Bruton and Samuda make the point that while learners seemed able to deal with apparent, immediate breakdowns in communication, several other, more subtle types of breakdown occurred which the students did not (and probably could not) treat. They suggest that learners be given an explanation of the various kinds of communication breakdowns that can occur, that they be taught strategies for coping with them, and that they be given explicit error monitoring tasks during group work.

Somewhat related to this work on error treatment is the analysis by Morrison and Low (1983) of monitoring in non-native discussions. Morrison and Low point out that their subjects, in addition to monitoring their own speech, self-correcting for lexis, syntax, discourse, and truth value without feedback from others and in a highly communicative context, also monitored the output of their interlocutors. This interactive view of monitoring, of making the struggle to communicate "a kind of team effort" (243), includes the kind of negotiation that Varonis and Gass are describing. The transcripts presented by Morrison and Low, however, show a wide divergence in the extent to which groups pay attention to and provide feedback on their members' speech. While some groups seemed to be involved in the topic and helped each other out at every lapse, other groups appeared totally absorbed in their own thoughts and inattentive to the speaker's struggles to communicate.

Quantity of practice: Students receive significantly more individual language practice opportunities in group work than in lockstep lessons (Long, Adams, McLean, and Castanos 1976, Doughty and Pica 1984, Pica and Doughty in press). They also receive significantly more practice opportunities in NNS/NNS than in NS/NNS dyads (Porter 1983), more when the other NNS has greater rather than equal proficiency in the SL (Porter 1983), and more in two-way than in one-way tasks (Doughty and Pica 1984).

Variety of practice: The range of language functions (rhetorical, pedagogic, and interpersonal) practiced by individual students is wider in group work than in lockstep teaching (Long, Adams, McLean, and Castanos 1976).

The fact that the level of accuracy maintained in unsupervised groups has been found to be as high as that in teacher- monitored, lockstep work should help to allay fears that lower quality is the price to be paid for higher quantity of practice. The same is true of the findings that monitoring and correction occur spontaneously (although variably) in group work and that it seems possible to improve both through student training in correction techniques, if that is thought desirable. The apparently spontaneous occurrence of other- correction probably diminishes the importance sometimes attached to designation of one student in each group as leader, with special responsibility for monitoring accuracy. However, group leaders may still be needed for other reasons, such as ensuring that a task is carried out in the manner the teacher or materials writer intended. (See Long 1977 for further details concerning the logistics of organizing group work in the classroom.)

For many teachers, of course, concern about errors occurring and/or going uncorrected has diminished in recent years, since second language acquisition research has shown errors to be an inevitable, even "healthy," part of language development. In fact, some teachers have been persuaded by theories of second language acquisition, such as Krashen's (1982) Monitor Theory, and/or by new teaching methods, such as the Natural Approach (Krashen and Terrell 1983), to focus exclusively on communicative language use from the very earliest stages of instruction. Many others, while not abandoning attention to form altogether, are eager to ensure that their lessons contain sizable portions of communication work, even though this will inevitably involve errors.

For such teachers, the most interesting findings of the research on interlanguage talk do not concern quantity and variety of language practice or accuracy and correction, but rather, the negotiation work in NNS/NNS conversation. The findings of each of five studies which have looked at the issue of whether learners can accomplish as much or more of this kind of practice working together as with a NS are very encouraging.

The related finding that students of mixed SL proficiencies tend to obtain more practice in negotiation than same-proficiency dyads suggests that when students with the same needs are working in small groups on the same materials or tasks, teachers of mixed-ability classes would do well to opt for heterogeneous (over homogeneous) ability grouping, unless additional considerations dictate otherwise. The fact that groups of mixed native language backgrounds tend to achieve greater amounts of negotiation also suggests grouping of students of mixed language backgrounds together where possible. For many teachers of multilingual classes, this would in any case be preferable, since it is one means of avoiding the development of "classroom dialects" intelligible only to speakers of a common first language--a phenomenon also avoidable through students having access to speakers of other target language varieties in lockstep work or outside the classroom.

The finding concerning mixed first language groups does not mean, of course, that group work will be unsuccessful in monolingual classrooms, which is the norm in many EFL situations. To reiterate, the research shows clearly that the kind of negotiation work of interest here is also very successfully obtained in groups of students of the same first language background. Things simply seem slightly better with mixed language groups.

Finally, the findings of research to date on interlanguage talk offer mixed evidence for the claimed advantages of two-way over oneway tasks in NS-NNS conversation. However, recent work on this issue seems to indicate that the claims are probably justified in the NNS-NNS context, too. Further, it appears to be the combination of small-group work (including pair work) with two-way tasks that is especially beneficial to learners in terms of the amount of talk produced, the amount of negotiation work produced, and the amount of comprehensible input obtained.

In this light, teachers might think it desirable to include as many two-way tasks as possible among the activities students carry out in small groups. It is obviously useful to have students work on oneway tasks, such as telling a story which the listener does not know or describing a picture which the listener attempts to draw on the basis of the description alone. However, because one participant starts with all the information in such tasks, the other group members have nothing to "bargain" with; this limits the ability of the latter to negotiate the way the conversation develops. (Some one-way tasks in fact become monologues rather than conversations.)

In conclusion, it should be remembered that group work is not a panacea. Teacher-fronted work is obviously useful for certain kinds of classroom activities, and poorly conceived or organized group work can be as ineffective as badly run lockstep lessons. Furthermore, additional information is still needed on such issues as the optimum size, composition, and internal organization of groups; about the structuring and management of tasks to be done in groups; and about the relationship between group work and teacber-led instruction.

Despite these caveats, the authors are encouraged by the initial findings of what we hope will develop into a coherent and cumulative line of classroom-oriented research: studies of interlanguage talk. Together with theoretical advances concerning the role of input in second language acquisition, the studies we have reviewed have already contributed a psycliolinguistic rationale to the existing pedagogical arguments for group work in the SL classroom.

Patricia A. Porter, Assistant Professor of English, coordinates the MA/TEFL Program and the ESL Program at San Francisco State University and teaches in both programs. Her research interests include interlanguage talk, and her most recent publication is an ESL oral communication text.

Doughty, Catherine, and Teresa Pica. 1984. Information gap tasks: do they facilitate second language acquisition? Paper presented at the 18th Annual TESOL Conference, Houston, March 1984.

Fanselow, John F. 1977. Beyond Rashomon- conceptualizing and describing the teaching act. TESOL Quarterly 11(l):17-39.

Fitz-Gibbon, C.I., and D.G. Reay. 1982. Peer-tutoring: brightening up FL teaching in an urban comprehensive school. British Journal of Language Teaching 20(l):39-44.

Flanders, Ned. 1970. Analyzing teaching behavior. Reading, Massachusetts: Addison-Wesley.

Gaies, Stephen J. 1983a. The investigation of language classroom processes. TESOL Quarterly 17(2): 205-217.

Gaies, Stephen J. 1983b. Learner feedback: an exploratory study of its role in the second language classroom. In Classroom oriented research in second language acquisition, Herbert W. Seliger and Michael H. Long (Eds.), 190- 213. Rowley, Massachusetts: Newbury House Publishers, Inc.

Gass, Susan, and Evangeline Marlos Varonis. In press. Negotiation of meaning in non-native sp eaker- non- native speaker conversation. In Input and second language acquisition, Susan Gass and Carolyn Madden (Eds.). Rowley, Massachusetts: Newbury House Publishers, Inc.

Hatch, Evelyn. 1983. Psycholinguistics: a second language perspective. Rowley, Massachusetts: Newbury House Publishers, Inc.

Hawkins, Barbara. In press. Is an "appropriate response" always so appropriate? In Input and second language acquisition, Susan Gass and Carolyn Madden (Eds.). Rowley, Massachusetts: Newbury House Publishers, Inc.

Hoetker, James, and William P. Ahlbrand. 1969. The persistence of the recitation. American Educational Research Journal 6(l):145-167.

Hunt, Kellogg W. 1970. Syntactic maturity in school children and adults. Monographs of the Society for Research in Child Development 53:1 (Serial No. 134).

Krashen, Stephen D. 1980. The input hypothesis. In Georgetown Roundtable on Languages and Linguistics, James E. Alatis (Ed.), 168-180. Washington, D.C.: Georgetown University Press.

Krashen, Stephen 1). 1982. Principles and practice in second language acquisition. Oxford: Pergamon Press.

Krashen, Stephen D., and Tracy Terrell. 1983. The natural approach: language acquisition in the classroom. New York: Pergamon Press.

Littlejohn, Andrew P. 1982. Teacherless language learning groups: an experiment. Manuscript, University of Lancaster.

Littlejohn, Andrew P. 1983. Increasing learner involvement in course management. TESOL Quarterly 17(4):595- 608.

Long, Michael H. 1975. Group work and communicative competence in the ESOL classroom. In On TESOL '75, Marina Burt and Heidi Dulay (Eds.), 211-233. Washington, D.C.: TESOL.

Long, Michael 11. 1977. Group work in the teaching and learning of English as a foreign language-problems and potential. English Language Teaching Journal 31(4):285-292.

Long, Michael H. 1980. Input, interaction and second language acquisition. Ph.D. dissertation, University of California, Los Angeles.

Long, Michael H. 1981. Input, interaction, and second language acquisition. In Native language and foreign language acquisition, Harris Winitz (Ed.). Annals of the New York Academy of Sciences 379:250-278.

Long, Michael H. 1983a. Linguistic and conversational adjustments to nonnative speakers. Studies in Second Language Acquisition 5(2):177-193.

Long, Michael H. 1983b. Native speaker/non- native speaker conversation in the second language classroom. In On TESOL'82, Mark A. Clarke and jean Handscombe (Eds.), 207-225. Washington, D.C.: TESOL.

Long, Michael H. In press. Input and second language acquisition theory. In Input and second language acquisition, Susan Gass and Carolyn Madden (Eds.). Rowley, Massachusetts: Newbury House Publishers, Inc.

Long, Michael H., Leslie Adams, Marilyn McLean, and Fernando Castaiios. 1976. Doing things with words: verbal interaction in lockstep and small group classroom situations. In On TESOL '76, Ruth Crymes and John Fanselow (Eds.), 137-153. Washington, D.C.: TESOL.

Long, Michael H., and Charlene Sato. 1983. Classroom foreigner talk discourse: forms and functions of teachers' questions. In Classroom oriented research in second language acquisition, Herbert NA'. Seliger and Michael H. Long (Eds.), 268-285. Rowley, Massachusetts: Newbury House Publishers, Inc.

McCurdy, Peggy L. 1980. Talking to foreigners: the role of rapport. Ph.D. dissertation, University of California, Berkeley.

Nlehan, Hugh. 1979. "What time is it, Denise?": asking known information questions in classroom discourse. Theory into Practice 18(4):285-294.

Morrison, Donald M., and Graham Low. 1983. Monitoring and second language learners. In Language and communication, Jack C. Richards and Richard NN', Schmidt (Eds.), 228-250. New York: Longman.

Pica, Teresa, and Catherine Doughty. In press. Input and interaction in the communicative language classroom: a comparison of teacber-fronted and group activities. In Input and second language acquisition, Susan Gass and Carolyn Madden (Eds.). Rowley, Massachusetts: Newbury House Publishers, Inc.

Porter, Patricia A. 1983. Variations in the conversations of adult learners of English as a function of the proficiency level of the participants. Ph.D. dissertation Stanford University.

Rowe, Mary Budd. 1974. Pausing phenomena: influence on the quality of instruction. Journal of Psych olingu istic Research 3(2):203-224.

Swain, Merrill. In press. Communicative competence: some roles of comprehensible input and comprehensible output in its development. In Input and second language acquisition, Susan Gass and Carolyn Madden (Eds.). Rowley, Massachusetts: Newbury House Publishers, Inc.

Tarone, Elaine E. 1981. Some thoughts on the notion of communication strategy. TESOL Quarterly 15(3):285-295.

Varonis, Evangeline Marlos, and Susan Cass. 1983. "Target language" input from non-native speakers. Paper presented at the 17th Annual TESOL Convention, Toronto, March 1983.