Output and Beyond to Dialogue: A Review of Merrill Swain's Current Approach to SLA

David J. Woodfield, Human International University

On the weekend of February 22nd and 23rd, Merrill Swain gave a seminar at Temple University in Osaka entitled "The Role of Output in Second Language Learning." A more accurate title for Dr. Swain's address though might be "Output and Beyond." First, Swain discussed the background to her interest in output. Then, she outlined three functions that she feels output has in facilitating language learning. Finally she moved beyond output alone to embed the concept in the idea of dialogue and then proceeded to show how,when language learners discuss their output in certain circumstances, what is actually going on at exactly that time is language learning. Swain's idea here, then, is not of dialogue as leading to language acquisition but of dialogue as the context in which language acquisition takes place online.

The Background to Swain's Interest in Output

If asked how language is acquired, many teachers would reply that it is through comprehensible input, through understanding messages in the L2 that are just a little above one's current language level. There is no doubt that Krashen's input hypothesis still holds great sway among language teachers. Indeed , it would be fair to say that the communicative task-based approach much used in classrooms is, to some extent, based on Krashen's theory. It seems intuitively true, however, that not only comprehension, but also production, has a direct role to play in acquiring a language.

Swain's interest in output evolved in the context of her ongoing study of Canada's French immersion programs. In such programs, non-French speaking children are placed as early as kindergarten into school classes where French is the sole language of instruction. Students are immersed in the L2 and are thus provided with what would seem to be a nearly ideal, acquisition-rich environment. Yet after doing their schooling in such an environment, do the students acquire native-like competency? Swain investigated the accuracy of predictions based on the comprehensible input hypothesis that immersion would help students acquire native-like competency.

Studies Swain (Swain, 1996) undertook in the eighties found that while immersion students acquired native-like reception skills they typically maintained certain non-target-like structures in their production. Swain wondered why, and therefore carefully examined what went on in immersion classrooms. She found that: the input itself was limited in various ways (for example, some grammatical structures such as conditional tenses were seldom used); although grammar was taught, it was taught discretely point by point and was seldom referred to during content area study; in their immersion classes, students actually produced language much less frequently than in classes taught in their L1; and when immersion students did produce, few of their linguistic errors were responded to, and when their errors were responded to, they were not responded to in any systematic way by the teachers.

As a result of her findings Swain then hypothesized that the encouragement of output--and in particular accurate output--is necessary for students to progress further towards target-like competence. Swain also finds support for an output hypothesis in the field of psycholinguistics, where researchers have theorized that the complete processing of syntax is not necessary to understand messages. Knowledge of context and lexical items can in some instances enable the understanding of the message content of an utterance without an understanding of its syntax. On the other hand, complete syntactic processing is necessary to produce "accurately" (i.e., like native speakers) because speakers must place vocabulary into a grammatical structure in creating sentences. Swain feels that in such a context the acquisition of structures is more likely, since attention must be paid to them.

The Three Functions of Output

To consider in more detail just how output provides opportunity for acquisition, let's now turn to the three functions of output that Swain discussed in her talk: noticing, hypothesis testing, and the reflective or metalinguistic function.

Noticing. Producing language causes learners to notice gaps in their linguistic knowledge. In other words, learners sometimes come to the realization that they don't know how to produce certain linguistic forms. Production, then, stimulates them "to notice what they do not know or know only partially"(Swain, 1995).

Hypothesis testing. Through noticing this gap in their knowledge, language learners may reanalyze their knowledge of the language system. On the basis of this analysis they then generate and test alternative ways of saying what they want to say. In other words they try to fill the gap in their knowledge. Noticing the gap thus stimulates hypothesis testing, the second function of output. Language learners can acquire new structures through this process, Swain asserted, or strengthen their control over previously acquired forms.

Reflective or metalinguistic function. Reflective or metalinguistic function refers to the fact that in trying to solve a problem in their output learners may consciously reflect upon the nature of the language system. Swain argues that such reflection can aid acquisition. Here hypotheses aren't simply generated and tested, but language is used to reflect upon the process. Swain argues that such reflection can aid acquisition in that it makes the process of noticing and hypothesis testing more explicit to the learner.


Before discussing the studies which provide support for Swain's characterization of the role of output, let's turn to her conceptualization of dialogue, the construct within which she places output in her most recent work.

In her appraisal of dialogue, Swain spent some time discussing the great socio-cultural theorist Vygotsky's ideas. Briefly, Vygotsky saw learning, and indeed human thought processes as a whole, as originating in interaction. He considered cognitive development as a matter of the appropriation of dialogue--the internalization of dialogue experienced during interaction with caregivers, for example.

Dialogue to Vygotsky, then, is a cognitive activity. Language, he believed, mediates between the individual and knowledge. It is language we put to good effect as a tool in attaining knowledge. Dialogue with others or internally is a significant context within which learning takes place.

According to Swain, output and its functions which facilitate language learning--the noticing of a gap in one's knowledge, hypothesis formation, testing and evaluation, and reflection upon language--can be profitably seen as existing within dialogue. Some of Swain's comments at the seminar reveal very clearly her movement to a position that dialogue is a cognitive activity in which language learning may be located:

  • "When I first thought about the metalinguistic function I saw it as talking about language, but now I see it as reflecting language learning."
  • "Language is a cognitive activity... in which is embedded actual language learning."
  • "Hypothesis testing is language learning."
  • "In dialogue we can see language learning in progress."

If dialogue is where learning takes place, the objective for the researcher and the teacher becomes to structure opportunities for dialogue about language. Thinking about the language you want to produce, and talking about that language become key activities where Swain expects to find language learning taking place. She sets up contexts where dialogue about language occurs and studies the language produced in those contexts, looking for evidence of language learning.

The Studies

In her seminar Swain discussed studies by herself and her colleagues at the Ontario Institute for Studies in Education of the University of Toronto, and some undertaken by other researchers, to research whether output or dialogue lead to language learning.

In a study published in 1995, Swain and Lapkin set out to assess the effects of output upon the "inner speech"of learners. Eighth grade immersion students were asked to assume the role of journalists, and to write a short article on an environmental problem, thinking aloud as they did so, thus allowing the researchers to see the impact of output upon the learners' thought processes. The researchers found that each of the students noticed, and responded to, a language problem in their output an average of just over ten times. In addition, they found that students analyzed their knowledge of the language in order to solve their problems. In one example from the study, a student who had written about how phosphates released into lakes cause plants to grow to such a great size that they kill the fish, struggled with how to say "kill the fish." She thought aloud as follows:

"et mort (and dies). I don't know. I don't know because mour. . . mourir les poissons (to die the fish), it's like mourir is something that you do . It's not something that someone does to you. So it's more like being murdered and not dying. So uhm . . . et tue toutes les poissons (and kills all the fish), or something like that."

Here the student notices a gap in her knowledge - she notices that there is something wrong with "mourir les poissons." Then she explicitly reflects on the nature of the language system: "mourir is something that you do. It's not something that someone does to you."As a result of her reflection, she selects an alternative form to express her meaning: "so uhm . . . et tue toutes les poissons, or something like that." The need to produce output, Swain suggested, has either helped the learner realize the distinction between transitive and intransitive verbs for the first time, or it has consolidated the learner's knowledge of this distinction, in that she has applied it to additional forms. Output led to noticing the gap, additional syntactic processing, the creation of a more accurate output, and the development of the student's interlanguage system in the process. The entire process can be considered as an example of language learning through private speech-an internal dialogue.

Other studies discussed by Swain focus upon the effects of dialogue between two or more participants, rather than within one learner. To structure contexts in which learners would communicate about language, Swain and other researchers turned to various kinds of group composing tasks. Swain's students and colleagues, Kowal (Kowal & Swain, 1997) and La Pierre (1994) used the dictogloss in their studies, Swain and Lapkin (in press) used a "picture jigsaw" story composition task, and Donato (1994) a pre-skit oral task. The selection of tasks is significant to the researchers in that those that stimulate dialogue about language and lead to language learning can be earmarked for further use in the classroom.

In her study, La Pierre had her students listen to a short narrative about a nightmare delivered at normal speed, take notes, and then reconstruct the passage in pairs as close as possible to the original -- a dictogloss task. She found that pairs for whom the researchers modeled the task and gave metalinguistic explanations for their choices of forms negotiated about language form an average of almost 15 times. In addition, pairs negotiated correct solutions for the language problems most of the time, and when doing so they identified the correct usage for the pertinent form 79% of the time in a post-test given a week later. This study seems to show that learners acquired forms when they successfully negotiated form and meaning (if correct identification in the post-test can be considered a measure of acquisition). A positive correlation between forms negotiated and post-test scores was also found by Swain and Lapkin in their "picture jigsaw study." Additionally, Donato's study in which he had groups of students make notes as part of their preparation for skits, 75% of the forms negotiated accurately were used accurately in the skits performed a week later. Swain sees these studies as support for the hypothesis that dialogue can lead to second language acquisition.

An example from the Donato study shows the effects of dialogue upon language learning that are typical of all of the above studies. In this study, the students create a skit in which a husband and wife are discussing a fur coat which the husband has bought for his mistress, the wife thinking it is her anniversary present. While creating the skit, the students produce and discuss their production; they speak about their language.

Speaker 1: . . . and then I'll say . . . tu as souvenu notre anniversaire de marriage . . . or should I say mon anniversaire?
Speaker 2: Tu as . . .
Speaker 3: Tu as . . .
Speaker 1: Tu as souvenu . . . 'you remembered?'
Speaker 3: Yeah but isn't that reflexive? Tu t'as . . .
Speaker 1: Ah, tu t'as souvenu.
Speaker 2: Oh, it's tu es
Speaker 3: tu es, tu es, tu . . .
Speaker 1: t'es, tu t'es
Speaker 3: tu t'es
Speaker 1: Tu t'es souvenu

Here the students collectively build a text, and in the process develop their knowledge of the language. They negotiate the structure of the phrase "you remembered" in French, going through four stages of discovery: a)tu as, b) t'as , c) tu es , d)tu t'es. First, all three speakers mull the initial phrase over in their minds, perhaps testing it against their intuition: "tu as . . . tu as . . . tu as souvenu . . . 'you remembered?'" Then speaker 3 suggests that the form might be reflexive-"tu t'as" and speaker 1 agrees. Then speaker 2 move the discussion up another step towards the target by noticing that the auxiliary should be etre not avoir - "Oh, it's tu es" and finally speaker 1 realizes that the new auxiliary also needs to be given in the reflexive form and so "tu t'es," the target structure, is achieved. Due to output, students notices a problem in output and follows a chain of hypothesis testing which leads to the resolution of the problem. Here the L1 is used to frame and assess the hypotheses-"or should I say . . . yeah, but isn't that reflexive? . . . ah," acting as a cognitive tool to achieve knowledge of the L2. To use Swain's words, through dialogue, the participants reach a deeper understanding of the language in context than they would have been able to reach individually. Through dialogue, knowledge is co-constructed and a degree of language learning takes place, right there and then.


Swain presented examples from the research where gender, tense, prepositions, vocabulary, and even register were negotiated by students in the context of the joint creation of text. Her seminar challenged its participants to reappraise their understanding of the language acquisition process and to consider the roles played by output and, in a wider sense, dialogue. Her seminar presented us with exciting possibilities for developing the accuracy of our students' production in a communicative context. This seminar has shown me that I need not be shy of having my students discuss language form, and I am open to the fact that such an activity, if structured appropriately, may be a fruitful context for language learning.


  • Donato, R. (1994). Collective scaffolding in second language learning. In J.P. Lantolf and G. Appel (Eds.), Vygotskian Approaches to Second Language Learning. Norwood, NJ: Ablex.
  • Kowal, M. and Swain, M. (1997). From semantic to syntactic processing: How can we promote metalinguistic awareness in the French immersion classroom? In R. K. Johnson & M. Swain (Eds.), Immersion Education: International Perspectives (pp. 284-309). Cambridge: Cambridge University Press.
  • La Pierre, D. (1994). Language output in a cooperative learning setting: Determining its effects on second language learning. Unpublished master's thesis, University of Toronto (OISE), Toronto, Canada.
  • Swain, M. (1995). Three functions of output in second language learning. In G. Cook and B. Seidlhofer (Eds.), Principle and practice in applied linguistics: Studies in honour of H.G. Widdowson (pp. 125-144). Oxford: Oxford University Press.
  • Swain, M. (1996). Integrating language and content in immersion classrooms: Research perspectives. The Canadian Modern Language Review, 52, 529-548.
  • Swain, M. & Lapkin, S. (1995). Problems in output and the cognitive processes they generate: A step towards second language learning. Applied Linguistics , 16, 371-391.
  • Swain, M. & Lapkin, S. (1997). Interaction and second language learning: Two adolescent French immersion students working together. Toronto: OISE/UT MS