Show Summary Details

Page of

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). (c) Oxford University Press, 2015. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in Oxford Handbooks Online for personal use (for details see Privacy Policy).

Subscriber: null; date: 23 August 2017

Interactionist Perspectives on Second Language Acquisition

Abstract and Keywords

This article analyses the idea of second language acquisition form an interactionist perspective. The field of second language acquisition has been studied from many angles. This broad scope is due in part to the myriad disciplinary backgrounds of scholars in the field. This article deals with the interactionist perspective and, as such, is primarily concerned with the environment in which second language learning takes place. It is important to note from the outset that this perspective is by and large neutral as to the role of innateness. In other words, it is compatible with a view of second language acquisition that posits an innate learning mechanism; it is also compatible with a model of learning that posits no such mechanism. This article deals with interactionist approaches focusing on how learners use their linguistic environment to build their knowledge of the second language. To summarize, the interaction approach considers, production of language as constructs that are important for understanding second language learning.

Keywords: perspectives, interactionist perspective, language acquisition, language learning, learning mechanism, linguistic environment

I. Introduction

The field of second language acquisition (SLA) has been studied from many angles. This broad scope is due in part to the myriad disciplinary backgrounds of scholars in the field. This chapter deals with the interactionist perspective and, as such, is primarily concerned with the environment in which second language learning takes place. It is important to note from the outset that this perspective is by and large neutral as to the role of innateness. In other words, it is compatible with a view of SLA that posits an innate learning mechanism (e.g., universal grammar [UG]); it is also compatible with a model of learning that posits no such mechanism. A word of caution is in order, however, as the situation is far more complex than these few terse statements suggest. It is not the case that interaction does not relate to issues of UG. Rather, what is intended is that the relationship of interaction to acquisition per se does not depend on whether there is or is not an innate mechanism that guides the learning of a second language. To provide an example of what the relationship between interaction and a UG account of learning might be, one notes that there are two kinds of evidence that are available to learners:1 positive evidence and negative evidence. Positive evidence refers to the language that a learner hears or reads and is clearly available through the linguistic environment; negative evidence is more complex. Negative evidence refers to information about what is incorrect in the language produced by a learner but not necessarily about what is needed to (p. 218) make an appropriate correction. Thus, overt correction is a form of negative evidence that not only indicates an error but also provides information about what is wrong. It may be that for complex syntax, negative evidence is necessary; it may also be that negative evidence is difficult to provide in other than an explicit (pedagogical) format. In other words, issues that form part of learnability theory (e.g., UG) may find answers in the conversational interactions in which learners engage in that these interactions provide the forum for both positive and negative evidence.

I have argued elsewhere (Gass, 1997) that the dichotomy between innatist and environmental approaches is ill-conceived in the sense that a presumed dichotomy leads only to a discussion of “which is correct” rather than of “how they complement one another.” But the goal of second language research must be the determination of how these approaches are intertwined, assuming that both are indeed relevant, even though they provide different sorts of explanations (see also Pinker, 1994). This chapter deals with interactionist approaches focusing on how learners use their linguistic environment (in particular, conversational interactions) to build their knowledge of the second language. To summarize briefly, the interaction approach considers exposure to language (input), production of language (output), and feedback on production (through interaction) as constructs that are important for understanding how second language learning takes place. Gass 2003 argues that interaction research “takes as its starting point the assumption that language learning is stimulated by communicative pressure and examines the relationship between communication and acquisition and the mechanisms (e.g., noticing, attention) that mediate between them” (p. 224). Long (1996) makes a similar claim, proposing that “environmental contributions to acquisition are mediated by selective attention and the learner's developing L2 processing capacity,” and that these resources “are brought together most usefully … during negotiation for meaning. Negative feedback obtained … may be facilitative of L2 development” (p. 414). A related construct, output was emphasized in Swain's (1993) research: “Learners need to be pushed to make use of their resources; they need to have their linguistic abilities stretched to their fullest; they need to reflect on their output and consider ways of modifying it to enhance comprehensibility, appropriateness, and accuracy” (pp. 160–161).

In what follows, I discuss the basic tenets of the interaction approach (see Gass and Mackey, 2007a) and then discuss learner-internal variables that impact how learners do and do not utilize language information stemming from interaction.

II. Input

Input has had an uneven history in the development of (second) language research. In behaviorist views of language learning, input was central to an understanding of how learners acquired a language, first or second. Imitation (p. 219) and habit formation were primary concepts in the acquisition process. If habits were formed through imitation, then it was necessary to examine the input to learners to determine what they were imitating. It was also necessary to examine the relationship between the input (what was to be imitated) and production (the product of imitation).

From the mid-1950s on, behaviorism was on the wane, and as a consequence so too was the importance ascribed to input. A new era of language research emerged, and acquisition research followed, deemphasizing the significance of input and focusing on the nature of the internal linguistic resources that a learner brings to the learning task.

By the early 1970s, scholars began to take a more balanced view of what was relevant and not relevant to the study of second language learning. With specific regard to input, Ferguson 1971, 1975) began to investigate the nature of input to nonproficient speakers of a language. In particular, he considered special registers such as “baby talk” (i.e., language addressed to young children) and “foreigner talk” (i.e., language addressed to nonnative speakers of a language). His focus was on the similarities between these two language systems, and his ultimate goal was understanding the nature of human language. With regard to foreigner talk, the system of particular interest to researchers in second language acquisition, certain common features became apparent. Speech directed towards nonproficient nonnative speakers was found to include speech modification ranging from phonological to syntactic. For example, speech tends to be slower, more clearly enunciated, and even louder. In terms of the lexicon, vocabulary tends to have a preponderance of common words. Syntax is simple, often including two sentences when one might normally expect to find one complex sentence. These characteristics are commonly found in speech to learners, although clearly there is variation among individuals.2

What function does modified speech serve? A lengthier discussion and exposé on this topic may be found in Gass 1997. For present purposes, it is important to point out that there are two perspectives from which one can answer this question. First is the perspective of the fluent speaker of the target language. It is likely that the purpose is to aid comprehension. One does what one can to ensure that one's conversational partner is able minimally to understand the general meaning and to be in a position to respond appropriately. The example below shows how a native speaker (NS), upon realizing that an original question may have been too difficult for a learner (NNS), modifies her speech to give the learner a greater opportunity to comprehend.

  • 1.

    NNS How have increasing food costs changed your eating habits?

    ➔NS Uh well that would I don't think they've changed 'em much right now, but the pressure's on.

    NNS Pardon me?

    ➔NS I don't think they've changed our eating habits much as of now…. (From Gass and Varonis, 1985)

(p. 220) In this example the reduced pronoun ('em ➔ them) is modified to include full information (our eating habits).

A second way of considering the function of modified speech is to examine the question from the perspective of the learner. Modified speech contributes to the likelihood that the learner can understand and can therefore get through what is essentially a social interaction. In other words, modified speech helps the learner participate in a conversation as fully as possible. Assuming that a learner is able to participate in a conversation, she is assured of receiving a greater quantity of input, can produce output, and can receive much-needed feedback on his/her production.

III. Interaction

The interactionist hypothesis—which was given initial prominence by Wagner-Gough and Hatch 1975 and refined by Long (1980, 1981, 1983) and others (Gass and Varonis, 1985, 1989; Mackey, 1999; Pica, 1987, 1988; Pica and Doughty, 1985; Pica, Doughty, and Young, 1986; Pica, Young, and Doughty, 1987; Schmidt and Frota, 1986; Varonis and Gass, 1985a)—has as its main claim that one route to second language learning is through conversational interaction (Gass, 1997; Long, 1996; Pica, 1994).

In Wagner-Gough and Hatch's original work, the role of conversation was valued, not just as a means for providing opportunities to practice previously learned language, but also as a locus of learning itself. Long, in his early work, showed how conversations involving nonfluent nonnative speakers of a language were quantitatively different from conversations in which both parties (assuming dyadic conversation) were equal and fluent participants. Wagner-Gough and Hatch 1975 and Long (1980, 1981, 1983) went beyond modified speech (e.g., simpler syntactic structures, easier vocabulary) to consider the structure of conversation itself. Following are examples of typical patterns found in conversations involving nonfluent, nonnative speakers of a language. This is not to say that these patterns do not exist in conversations involving fluent speakers—only that they are more frequent in nonnative learner speech.

  1. 2. Confirmation Check

    NNS c'è una verdi, uh … there is a green, uh.

    NS una verdi? a green? (From Mackey, Gass, and McDonough, 2000)

In this example from an English-speaking learner of Italian, the NS's questioning the word verdi, which in fact is inappropriate in Italian, resulted in a subsequent negotiation until the correct word pianta (plant) was recognized later in the exchange.

(p. 221) Examples 3–7 are crucial to an understanding of the interaction hypothesis and to an understanding of how modified interactions contribute to learning. In these examples, one can see various means of modifying a conversational structure, resulting in a greater likelihood of comprehension.

  1. 3. Comprehension Check

    NNS1 And your family have some ingress.

    NNS2 yes ah, OK, OK?

    ➔ NNS1 more or less OK? (From Varonis and Gass, 1985a)

  2. 4. Clarification request

    NS there's there's just a couple more things

    NNS a sorry? Couple?

    NS couple more things in the room only just a couple

    ➔ NNS couple? What does it mean couple? (From Mackey and Philp, 1998)

  3. 5. Or-choice

    NS Where did you go yesterday?

    NNS What?

    ➔NS Did you go to the zoo or to the garden? (Original data)

  4. 6. Topic-focused

    NS Did your friend travel with you to Italy and Switzerland?

    NNS What?

    ➔ NS Your friend, did she travel with you? (Original data)

  5. 7. Elaboration

    NNS How have increasing food costs changed your eating habits?

    NS Well, we don't eat as much beef as we used to. We eat more chicken and uh, pork, and uh, fish, things like that.

    NNS Pardon me?

    ➔ NS We don't eat as much beef as we used to. We eat more chicken and uh, uh pork and fish…. We don't eat beef very often. We don't have steak like we used to. (From Gass and Varonis, 1985)

Long (1996) defined the interaction hypothesis as

negotiation for meaning, and especially negotiation work that triggers interactional adjustments by the NS or more competent interlocutor, facilitates acquisition (p. 222) because it connects input, internal learner capacities, particularly selective attention, and output in productive ways. (pp. 451–452)

In addition,

it is proposed that environmental contributions to acquisition are mediated by selective attention and the learner's developing L2 processing capacity, and that these resources are brought together most usefully, although not exclusively, during negotiation for meaning. Negative feedback obtained during negotiation work or elsewhere may be facilitative of L2 development, at least for vocabulary, morphology, and language-specific syntax, and essential for learning certain specifiable L1-L2 contrasts. (p. 414)

It is through negotiation that the learner may direct attention to an area of the target language (1) about which she may be entertaining an hypothesis (or about which she is trying to formulate a hypothesis) or (2) about which she has no information. This is not to say that learning necessarily takes place during a conversation; the interaction itself may only be the first step in recognizing that there is something to learn. Interaction may be a priming device representing the setting of the stage for learning. The following examples represent two ways in which conversation can lead to learning: (1) on the spot learning and (2) delayed learning.

Example 8 comes from a study that used stimulated recall (see Gass and Mackey, 2000) to determine what a learner was thinking during a prior interaction. The relevant part of the interaction is provided, followed immediately by the learner's comments on the interaction.

  1. 8. On the Spot Learning

    NNS so the people make a line in front of the place

    NS they are standing in line?

    ➔ NNS ahh, they are standing in line (Twelve turns later)

    ➔ NNS Beside the people standing in line (From Mackey, Gass, and McDonough, 2000)

Learner's retrospective comment:

Make a line is the same meaning as stand in line, actually I thought of this situation as make a line but after she said standing in line, her expression is better than mine so I changed mine.

The learner in this instance used the conversation as a way of obtaining new linguistic information, as is evidenced by her retrospective comments. She then tried out her newly learned information later in the conversation.

But sometimes learning does not take place immediately; time to “digest” the new linguistic input is often needed. This is illustrated in 9, which took place over the time period of a class.

  1. (p. 223) 9. Delayed Learning

    ➔NNS1 Kutsu o shiroi no kutsu o hamete imasu ka? Shoes ACC white (GEN [ sic ] shoes ACC wearing [ sic ] INT? Is the person wearing white shoes” ((wrong verb for “wear,” underlined))

    NNS2 Iie haite imasen. Ano::: No, he's not. Uh:: (correct verb for “wear,” underlined))

    ➔NNS1 Ano: ano: kuroi no kutsu o hai- hamete imasu ka? U:h u:h black (GEN [ sic ] shoes ACC wea-wearing [ sic ] INT? U'h u”h is the person wearing black shoes ((wrong word for “wear,” underlined))

    NNS2 Hai haite imasu Yes, he is ((correct verb for “wear,” underlined))

    NNS1 Anoo uh jiinzu o ::: jeans? Jinzu o ha uh :::: haite imasu ka? U:h uh: you're um jeans? Are you u :::h wearing Jeans?

    NNS1 Sorekara shiroi no kutsu ga kutsu o how do you say it? Kutso o ::: And white shoes shoes how do you say it for shoes?

    NNS3 haitemasu. haki-Wearing. Wear-

    ➔NNS1 what? Haite imasu ka? What? Is it “haite imasu?”

    ➔NNS4 Hanjiinzu o : : um oh what is it? Hare- haite imasu ka? Shorts ((sic)) ACC um oh what is it? Do you we-wear shorts?

    NNS4 (inaudible)

    ➔NNS1 Sorekara ::: kiroi to shiroi no kutsu g- o ::: yeah. Haite imasu ka? And ::: are you wearing yellow and white shoes? (From Ohta, 2001)

In this example the learner initially uses the incorrect form for the verb wear, haimete, rather than the correct form, haite. She begins to receive input with the correct form and eventually asks about the correct form, but it isn't until quite a bit later in the class that she finally, presumably as a result of the interaction, begins to use the correct form.

Learners do not always use a conversation to obtain new information; there are times when she or he uses it to test certain hypotheses about the second language and to receive feedback on production. Example 10 below shows how a learner uses a conversation to test a hypothesis:

  • 10.

    NNS poi un bicchiere then a glass

    INT un che, come a what, what?

    NNS bicchiere Glass (From research reported on in Mackey, Gass, and McDonough, 2000)

(p. 224) Learner's retrospective comment:

I was drawing a blank. Then I thought of a vase but then I thought that since there was no flowers, maybe it was just a big glass. So, then I thought I'll say it and see. Then, when she said “come” [what] I knew that it was completely wrong.

In this instance, the learner is throwing out a word to see where it gets her.

3.1 Recasts

Feedback to learners can be explicit (e.g., an overt correction) or implicit, as often illustrated through negotiation work. Perhaps the most subtle type of feedback comes in the form of a recast, a reformulation of an incorrect utterance while maintaining the original meaning of the utterance. An example is given in 11 in which the native speaker recasts the nonnative speaker's incorrect question:

  • 11.

    NNS I think some this girl have birthday and and its big celebrate

    ➔NS big celebration

    NNS oh (From Mackey and Philip, 1998)

A question arises as to the effectiveness of recasts. Are they noticed, and, if so, are they taken as a form of correction? Lyster and Ranta 1997 collected data from grades 4–6 children in French immersion programs. They were primarily concerned with the reaction by the student immediately following a recast inasmuch as this reveals what the student does with the feedback. Despite the preponderance of recasts in their database, recasts were not found to have an impact on subsequent production. Using the same database, Lyster 1998 found that there was some confusion between the corrective and approval functions of recasts, thereby questioning their usefulness in terms of corrective feedback.

Other studies, however, do show a positive effect for recasts (for reviews, see Long, 2007, and Nicholas, Lightbown, and Spada, 2001). Mackey and Philp 1998 argue that using the production immediately following a recast may not be the most appropriate way to determine effectiveness. They make the point that if one is to consider effectiveness (i.e., development/acquisition), then one should more appropriately measure delayed effects (see, e.g., Gass, 1997; Gass and Varonis, 1994; Lightbown, 1998). The study on the acquisition of English questions (Mackey and Philp, 1998) showed that for more advanced learners, recasts plus negotiation were more beneficial than negotiation alone. This was the case even though there was not always evidence for a reaction by the learner in the subsequent turn (see Oliver, 1995). Thus, the effectiveness of recasts may be in part dependent on developmental stages (see, e.g., Iwashita, 2003; Philp, 2003). Further, as Egi (2007) has shown, the value of recasts may be dependent on their interpretation—in other words, are they thought to be a comment on content (a practice that is not effective), or are they taken to be a form of positive and/or negative evidence (in which case they appear to be of some benefit)?

(p. 225) Recent meta-analyses (Mackey and Goo, 2007; Russell and Spada, 2006) have addressed the effects of recasts. Both studies reported large effect sizes for implicit feedback (including recasts), but both also report that their data are based on very few studies given their inclusion criteria for their meta-analyses. Mackey and Goo also note that although there are strong effects on learning from recasts, there is little data on longer term effects.

3.2 Does Interaction Contribute to Learning?

In recent years, the question has moved from the arena of speculation to serious investigation of the effects of interaction. Three early studies led the way in this regard:

  1. 1. Loschky 1994, in an investigation of English learners of Japanese, found that interaction had a positive effect on comprehension but did not find an effect on the acquisition of vocabulary or on the acquisition of morphosyntax (locative expressions; see also Ellis, Tanaka, and Yamazaki (1994).

  2. 2. Gass and Varonis 1994 did show that the effects of interaction went beyond a subsequent turn. As in previous studies, vocabulary did not appear to be affected, although general discourse organizational strategies were affected.

  3. 3. Mackey 1999 provides the most detailed support in favor of the effects of interaction. In her investigation of a single grammatical structure, question formation, she found a relationship between conversational interaction and development in that those who were involved in structure-focused interaction moved along a developmental continuum more rapidly than those who did not. Mackey's study supports the notion of interaction not necessarily (or not always) being the locus of immediate learning but often being the catalyst for later learning. She found that for the developmentally advanced structures, the effects of interaction were noted in delayed posttests rather than immediately.

In the past 30 years or so, numerous studies have been conducted on the relationship between interaction and learning, most showing a positive relationship (e.g., Adams, 2007; Trofimovich, Ammar and Gatbonton, 2007; Carpenter, Jeon, MacGregor, and Mackey, 2006; Ellis, 2007; Ellis, Loewen, and Erlam, 2006; Gass and Alvarez-Torres, 2005; Gass and Lewis, 2007; Jeon, 2007; Kim and Han, 2007; Loewen, 2005; Loewen and Nabei, 2007; Loewen and Philp, 2006; Lyster and Mori, 2006; Mackey, 2006; Mackey and Silver, 2005; McDonough, 2005, 2006, 2007; McDonough and Mackey, 2006; Pica, Kang, and Sauro, 2006; Sachs and Suh, 2007; Sagarra, 2007; Sato and Lyster, 2007; Sheen, 2007; Tarone and Bigelow, 2007; Tocalli-Beller and Swain, 2007). Only studies published since 2005 have been included due to limitations of space; for a more complete listing, see Mackey (2007).

In their meta-analysis, Mackey and Goo (2007) found that interaction is facilitative of the acquisition of both vocabulary and grammar. There is a stronger (p. 226) immediate effect for vocabulary, but a delayed and lasting effect on grammar. Both feedback and modified output were significant factors in promoting learning, but Mackey and Goo recommend a more nuanced investigation of these interactional components.

There is little doubt as to the facilitative effects of interaction on learning, but many questions remain unanswered. Some of the questions are descriptive, as was previously alluded to; that is, what types of feedback contribute to learning and under what circumstances? The second type of question is explanatory; that is, why does interaction contribute to learning? There are four areas that will be examined next:

  1. a. Attention

  2. b. Working memory (including inhibitory control)

  3. c. Output

  4. d. Contrast

IV. Attention

Central to the interaction hypothesis is the concept of attention or noticing. If interaction is to have an effect (either through negotiation or recasts), the learner must notice that his/her conversational partner is explicitly or implicitly making a correction.3 If there is no attention to a particular part of language during an interaction, then it is difficult to attribute the source of change to the interaction itself. Example 12 illustrates how the direct questioning of an utterance makes the learner notice the discrepancy between her pronunciation of yellow and the native speaker interviewer's pronunciation:

  • 12.

    NNS The color is /wellow/

    NS Is what?

    NNS /wellow/ /wellow/ color

    NS Yellow?

    NNS Yellow. (From Mackey, Gass and McDonough, 2000)

Learner's retrospective comment:

My pronounce is different. I say /wellow/ but yellow is the exact pronounce. Yellow, yellow.

The question remains as to what learners do notice. J. Williams 1999 considered learner-generated attention to form. She points out that most of the focus on form research within a classroom context focuses on teacher-generated attention. (See also Sharwood Smith's [1991, 1993] discussion of enhanced input and internally and (p. 227) externally induced salience). Williams found that learners are indeed capable of focusing attention to language form, but that there is variation according to proficiency level and even to activity types.

Mackey, Gass, and McDonough (2000) also investigated what learners notice in an interaction. Their investigation differed from that of Williams in that they were concerned with interactional feedback and how learners actually interpreted that feedback. Through a postinteraction stimulated recall of learners of English and learners of Italian, they found that learners do notice interactional feedback, but they do not do so in a uniform manner. Lexis and phonology are more likely to be noticed than aspects of morphosyntax. The results also suggest that the manner of feedback (e.g., recasts or negotiation) have different effects when occurring alone or in combination. A related study of heritage and nonheritage learners (Gass and Lewis, 2007) found that both nonheritage language learners and heritage language learners perceive phonological and lexical feedback much more accurately than morphosyntactic feedback, as had been observed in Mackey, Gass, and McDonough (2000). However, differences were noted in the area of semantic feedback, with nonheritage language learners generally not accurate in their perceptions about semantic feedback, whereas heritage language learners were. Neither of these studies investigated the next step; that is, the determination of what happens after learners notice a gap between their knowledge of the second language and the second language itself.

V. Working Memory

Within an interactive context, learners produce language, receive feedback, and, in an ideal situation, use the information contained in the feedback in a productive way. In so doing, a learner must notice the feedback given, determine what is relevant, and retain that information long enough to identify the precise part of language that is being corrected. The question remains as to why some individuals are more successful at this than others. One possible explanation relates to working memory capacity. Why some learners are able to focus attention on certain parts of an interaction better than others may be related to their ability to regulate their focus of attention, either through selectively attending to some part of language or by inhibiting others. Working memory is generally considered to incorporate both processing and storage functions of memory. Miyake and Shah 1999 define working memory as “those mechanisms or processes that are involved in the control, regulation, and active maintenance of task-relevant information in the service of complex cognition, including novel as well as familiar, skilled tasks” (p. 450). In other words, following work by Baddeley and Hitch (1974), working memory keeps representations in temporary storage, allowing operations on those representations to take place (Caplan, Waters, and Dede, 2007).

There have not been many studies that relate interactional success and working memory capacity. One of the early ones was by Philp 2003, who studied noticing. (p. 228) She did not measure working memory capacity, but did point the way to research involving working memory when she suggested that attentional resources may be at the base of understanding when learners do notice differences between their own utterances and those of the target language. Two studies by Sagarra (2007, 2008) indirectly looked at working memory capacity and interaction. In one—a computer-based study of feedback—Sagarra (2007) found that working memory capacity predicted learners' ability to benefit from recasts. In the second, Sagarra (2008) found that redundant grammatical information is not processed by low proficiency L2 learners (English as an L1; Spanish as an L2) with low working memory capacity.

In the most directly relevant study, Mackey, Philp, Egi, Fujii, and Tatsumi (2002) investigated the relationship among individual differences in verbal working memory, noticing of interactional feedback, and the L2 development of English question formation. Their study compared the interactional benefits of two groups of L2 learners (L1 Japanese; L2 English): those with low working memory capacity and those with high working memory capacity. The former group benefitted immediately from interaction, but the results did not persist on a delayed posttest (2 weeks after the treatment). Those with higher working memory capacity demonstrated more lasting benefits from communicative interaction as demonstrated in the delayed posttest. A possible explanation is that learners with higher working memory capacities engage in cognitive comparisons between target language forms and their own versions of the forms, impacting processing loads and immediate performance. Learners with lower working memory capacities, in contrast, may be better equipped to engage in immediate modifications to output, at a potential longer term cost to comparison, storage, and subsequent retrieval mechanisms. Thus, the emphasis for high-working memory capacity individuals may be on processing, whereas the emphasis for low-working memory capacity individuals may be on storage (for differing results, cf. Trofimovich, Ammar, and Gatbonton, 2007).

Initial analysis in a study of English-speaking learners of Italian (Gass, Behney, and Uzum, in preparation) suggests that more significant than working memory differences are differences in inhibitory control (a construct related to working memory)—that is, the ability to suppress information that is not relevant.

In sum, second language learners are exposed to more input than they can process; consequently, they must have a mechanism that enables them to sort through that input (see, e.g., Gass, 1997) to determine what is (momentarily) relevant and what is not. Working memory capacity may be one such mechanism and inhibitory control may be another.

VI. Output

In preceding sections, the nature of language that is directed toward learners and the function of modified language and/or modified conversational structure have been considered. However, in any discussion of the interaction hypothesis, there is (p. 229) a third prong to examine, and that is the role of output. In earlier conceptualizations of second language acquisition, output served little learning purpose, other than, perhaps, to reinforce previously learned linguistic knowledge. Swain's (1985) pioneering work (see also Swain, 1995, 2005) in this area came from observations of immersion programs in Canada (see also Kowal and Swain, 1997; Swain and Lapkin, 1995, 1998). She noted that children who had spent many years in immersion programs were still lagging in target-language-like abilities. In looking more carefully at the classroom context in which the target language was used and in which it was the prime source of information about the target language for these children, she noted that what was lacking was consistent and frequent use of the second language. She proposed that one needed more than input; learning a second language required a significant amount of output. Output, or language production, forces learners to focus on the syntax of an utterance and, consequently, on formulating hypotheses about how the target language works. This is different from receiving input because input involves primarily comprehension and comprehension often requires little syntactic organization. As a result, Swain introduced the notion of comprehensible output or “pushed” output.

Comprehensible output refers to the need for a learner to be “pushed toward the delivery of a message that is not only conveyed, but that is conveyed precisely, coherently, and appropriately” (Swain, 1985: 249). In a more recent explication of the concept, Swain claimed output may stimulate learners to move from the semantic, open-ended, nondeterministic, strategic processing prevalent in comprehension to the complete grammatical processing needed for accurate production. Output, thus, would seem to have a potentially significant role in the development of syntax and morphology (1995: 128).

Mackey 2002 conducted a study in which learners reflected on a previous interaction through a stimulated recall procedure. Example 13 provides insight into what the learner was thinking as she was engaged in a conversation.

  1. 13. Example of Pushed Output

    NNS: And in hand in hand [ sic ] have a bigger glass to see.

    NS: It's err. You mean, something in his hand?

    NNS: Like spectacle. For older person.

    NS: Mmmm, sorry I don't follow, it's what?

    NNS: In hand have he have has a glass for looking through for make the print bigger to see, to see the print, for magnify.

    NS: He has some glasses?

    NNS: Magnify glasses he has magnifying glass.

    NS: Oh aha I see a magnifying glass, right that's a good one, ok. (Mackey 2002)

Retrospective comment:

In this example I see I have to manage myerrerr [sic] expression because he does not understand me and I cannot think of exact word right then. I am thinking thinking [sic] it is nearly in my mind, thinking bigger and magnificate and (p. 230) eventually magnify. I know I see this word before but so I am sort of talking around around [sic] this word but he is forcing me to think harder, think harder for the correct word to give him so he can understand and so I was trying. I carry on talking until finally I get it, and when I say it, then he understand it, me [emphasis mine].

As can be seen, the learner was pushed (note the word forcing) through the negotiation sequences to make her language clearer.4

McDonough 2005 tested the output hypothesis directly in her study of Thai learners of English. In a study investigating the acquisition of English questions, four groups carried out communicative tasks. The groups focused on salience (enhancement) and opportunity to modify following feedback. She found that the best predictor of acquisition lies in the opportunity to modify one's speech. A later study (McDonough and Mackey, 2006), however, noted that mere repetition of an interlocutor's form (in their case, recasts) did not impact learning, whereas primed production (use of a form in the recast in later production) did.

Output, in sum, is important for a number of reasons, including forcing a learner to produce language about which feedback may often be given. However, precisely how it benefits learning is still open for investigation. Further, as noted by Sato and Lyster 2007, the interlocutor partner (native speaker or another learner) may impact the amount of modified output; learners tend to produce more when interacting with other learners than with a native speaker (see also Varonis and Gass, 1985a).

VII. Theory of Contrast

What sort of mechanism allows for learning to take place as a result of negative evidence derived from conversational interaction? One possibility to account for learning through conversation is the direct contrast hypothesis (Saxton, 1997), defined within the context of child language acquisition as follows:

When the child produces an utterance containing an erroneous form, which is responded to immediately with an utterance containing the correct adult alternative to the erroneous form (i.e., when negative evidence is supplied), the child may perceive the adult form as being in CONTRAST [emphasis in original] with the equivalent child form. Cognizance of a relevant contrast can then form the basis for perceiving the adult form as a correct alternative to the child form. (1997: 155)

Attention alone is not sufficient. A contrast must be attended to, or in SLA parlance, a gap must be noticed. And conversation provides a forum for the contrast to be detected, especially when the erroneous form and a correct one are in immediate juxtaposition.

(p. 231) VIII. Conclusion

It is likely that there are limitations to what can and cannot be learned through negative evidence provided through conversation. One possibility is that surface-level phenomena can be learned but abstractions cannot. This is consistent with Truscott's 1998 claim that competence is not affected by noticing. Negative evidence can probably not apply to long stretches of speech, given memory limitations (see Philp, 2003). But it may be effective with low-level phenomena such as pronunciation or the basic meanings of lexical items. In fact, these are precisely the areas that Mackey, Gass, and McDonough (2000) have isolated as those that are sensitive to feedback. Future research will need to determine the long-term effects of interaction on different parts of language (see also Gass, Svetics, and Lemelin, 2003; Trofimovic, Ammar, and Gatbonton, 2007).

Research on interaction began descriptively. In the past decade, as it became clear through empirical research that there was a positive relationship between interaction and learning, research moved toward an explanation of how this relationship works: What individual factors can help to explain differential benefits? In looking toward future research, individual cognitive differences will undoubtedly lead the way in attempting to answer this question, but other factors such as motivation, learning strategies, language aptitude, cognitive styles, learning context (e.g., second versus foreign language learning), language background (heritage versus nonheritage language learners), and social context will also likely have a role.

Notes:

(1.) There is a third type of evidence, indirect negative evidence (see Plough, 1994, 1995). To simplify matters, we do not deal with this complex type of evidence.

(2.) It is beyond the scope of this paper to comment on whether these systems are learned or not. It is important to note, however, that the extent to which an individual adopts foreigner talk characteristics in her speech may depend to an individual's experience with nonnative speakers. Varonis and Gass (1985b) describe an interaction in which a salesperson adopts very few foreigner talk features (she speaks rapidly, uses idioms, and uses anaphoric pronouns with no obvious referent). It was speculated that this is so precisely because of her lack of experience with noncomprehending individuals.

(3.) This is not to say that attention cannot come from a learner noticing a new form on her own (i.e., without an interlocutor's correction).

(4.) It is not clear from the description in the original article whether the learner had actually seen the phrase magnifying glass as part of the input (although she does say that she had seen the word [sic] before) and was trying to recall it or if she was generating it from what she had heard in the exchange. Regardless, it is through the interaction that this learner was able to come up with the correct word.