TESTING THE OUTPUT HYPOTHESIS - Shinichi Izumi, Martha Bigelow, Miho Fuj

SSLA, 21, 421–452. Printed in the United States of America. TESTING THE OUTPUT HYPOTHESIS Effects of Output on Noticing and Second Language Acquisition Shinichi Izumi, Martha Bigelow, Miho Fujiwara, and Sarah Fearnow Georgetown University This study addresses one of the functions of output proposed by Swain (1993, 1995, 1998). In particular, the activity of producing the target language may, under certain circumstances, prompt L2 learn- ers to recognize some of their linguistic problems and bring to their attention something they need to discover about their L2. Two re- search questions were posed: (a) Does output promote noticing of linguistic form? and (b) Does output result in improved performance on the target form? In treatment phase 1, participants reconstructed a short passage after being exposed to it, followed by a second exposure to the same input material and a second reconstruction opportunity. In phase 2, participants wrote on given topics, followed by the presentation of a model written by a native speaker. Participants wrote a second time on the same topic. To test the noticing function of output, participants underlined parts of the sentences they thought were “particularly necessary” for subsequent (re)production. The con- trol group was exposed to the same input materials but was not The study reported in this paper was conducted as a project for a seminar in second language acqui- sition at Georgetown University. An earlier version of this paper was presented at SLRF ’97 (Michigan State University) by Shinichi Izumi and Miho Fujiwara. The authors would like to thank Catherine Doughty for her assistance throughout the entire process of the study. Thanks are also due to Mi- chael Long, Michael O’Neill, Deborah Mifflin, and Christine Glover for their many insightful sugges- tions and comments. We would like to thank the students who participated in this study and the ESL instructors who allowed us to pilot tasks and tests in their classes. Thanks are also due to Yoshiho Shibuya and Takae Tsujioka for drawing the pictures needed for the tests and tasks. Finally, the authors thank the anonymous SSLA reviewers for their many helpful suggestions. Address correspondence to: Shinichi Izumi and Martha Bigelow, Linguistics Department, George- town University, Washington, DC 20057, e-mail: izumis@gusun.georgetown.edu and bigelowm@gusun. georgetown.edu.  1999 Cambridge University Press 0272-2631/99 $9.50 421 422 Shinichi Izumi et al. required to produce any output. Although phase 1 tasks resulted in noticing and immediate incorporation of the target form, the posttest performance failed to reveal their effects. In contrast, phase 2 tasks resulted in improvement on posttest 2. The results are discussed in terms of the efficacy of output in promoting noticing and learning and the conditions that may be required for output to be useful for SLA. In Canadian immersion programs, teachers provide students with a rich source of comprehensible input. Decades of research on these programs have shown that they are among the most successful large-scale L2 programs in existence. However, over the years, research on these programs has accumu- lated evidence indicating that merely providing large amounts of comprehen- sible input is not sufficient for the L2 learner to attain a high level of L2 proficiency (Hammerly, 1987; Harley, 1993; Harley & Swain, 1978). Findings from such studies have shown that L2 learners in immersion classes often demonstrate weaknesses in grammatical accuracy, despite gaining high-level listening comprehension skills and communicative fluency. Swain (1985) argued that one important reason why immersion learners display numerous grammatical errors in their L2 is that they actually engage in a small amount of language production. This characteristic of their lan- guage-learning environment may prevent them from going beyond their cur- rent level of L2 proficiency. Observational studies of interaction in French immersion classrooms have described typical examples of teacher-student in- teractions in which teachers did most of the talking and few of the utterances made by students were longer than a single clause (Allen, Swain, Harley, & Cummins, 1990; Swain, 1985). This kind of interaction permits students to op- erate successfully with their incomplete knowledge of the language because they are rarely pushed to be more accurate. Observations such as these have led Swain to conclude that “comprehensi- ble input,” although invaluable to the acquisition process, is not sufficient for learners to fully develop their L2 proficiency. Specifically, she argues that what students need is not only comprehensible input but also “comprehensi- ble output” if they are to be both fluent and accurate in the target language. The construct of comprehensible output posits that, when learners experi- ence communication difficulties, they will be pushed into making their output more precise, coherent, and accurate. According to Swain, “producing the tar- get language may be the trigger that forces the learner to pay attention to the means of expression needed in order to successfully convey his or her own intended meaning” (1985, p. 249). Since the Output Hypothesis was first proposed, Swain (1993, 1995) has ex- tended the scope of the hypothesis and identified the following three func- tions of output. First, output has a hypothesis-testing function. Producing output is potentially a way of testing one’s hypothesis about the target lan- Effects of Output on Noticing and SLA 423 guage. Learners can test comprehensibility and linguistic well-formedness of their interlanguage (IL) against feedback obtained from their interlocutors. This function of output relates directly to the notion of comprehensible out- put that Swain claimed to be necessary for immersion students. By producing output, learners can test their hypotheses and, by being pushed in the pro- cess of negotiation of meaning, they can be more accurate in their production. Second, output may have a metalinguistic function. It is claimed that “as learners reflect upon their own target language use, their output serves a met- alinguistic function, enabling them to control and internalize linguistic knowl- edge” (Swain, 1995, p. 126). In other words, output may cause the learner to engage in more syntactic processing than is necessary for the comprehension of input. This syntactic processing may lead to modified or reprocessed out- put—a possible step toward language acquisition. Reflection on language may not necessarily involve metalinguistic terminology. “Metatalk” produced with or without metalinguistic terminology may, according to Swain, serve the function of deepening the students’ awareness of forms, rules, and form-func- tion relationships, provided that the contexts are such that the language pro- duced by the learners serves some genuine communicative function. Finally, output may serve a “noticing/triggering” (or consciousness-raising) function. That is to say: [I]n producing the target language (vocally or subvocally) learners may no- tice a gap between what they want to say and what they can say, leading them to recognize what they do not know, or know only partially, about the target language. In other words, under some circumstances, the activ- ity of producing the target language may prompt second language learners to consciously recognize some of their linguistic problems; it may bring to their attention something they need to discover about their L2. (Swain, 1995, pp. 125–126) This function of output is consistent with the claim made by Schmidt and Frota (1986) that “a second language learner will begin to acquire the target- like form if and only if it is present in comprehended input and ‘noticed’ in the normal sense of the word, that is consciously” (p. 311). The role of output here is proposed to facilitate the process of noticing of both problems in one’s IL and the relevant features in the input. This noticing will then stimulate the processes of language acquisition. It is important to note that Swain does not claim that output is the only source of L2 acquisition. Her claim, rather, is that output can, under some conditions, promote language acquisition by allowing learners to recognize problems in their IL capabilities. Recognition of problems may occur because of either internal feedback (the very process of producing output may prompt syntactic processing of language, not just semantic or pragmatic processing, which is often invoked in processing language for the purpose of comprehen- sion) or external feedback (feedback obtained from interlocutors, teachers, etc., which informs the learners as to the well-formedness of their IL produc- 424 Shinichi Izumi et al. tion). This recognition of problems is believed to prompt learners to generate alternatives by searching their knowledge, assessing alternatives, applying ex- isting knowledge to known contexts or new contexts, and then using the re- sulting knowledge. Alternatively, recognition of problems through output may prompt learners to seek out relevant input with more focused attention (Swain & Lapkin, 1995). Several studies have been conducted to test the first and second functions of output—namely, the hypothesis-testing function and the metalinguistic function. The results from the studies related to the hypothesis-testing func- tion of output (Nobuyoshi & Ellis, 1993; Pica, 1988, 1992; Pica, Holliday, Lewis, & Morgenthaler, 1989; Takashima, 1994) show that learners often mod- ify their output in response to the linguistic demands of NS signals of compre- hension difficulty and also suggest that pushing learners to produce more comprehensible output may have a long-term effect. Results of the studies fo- cusing on the metalinguistic function of output (Donato, 1994; LaPierre, 1994; Swain, 1995) lend some support to the claim that producing language and re- flecting on it in an attempt to create meaning has positive effects on language- learning processes. In contrast to the hypothesis-testing function and the metalinguistic func- tion of output, there has been little research done to test the third hypothe- sized function of output—the noticing function. One of the few studies that has investigated this function of output is that of Swain and Lapkin (1995). They examined whether output would lead to conscious recognition of prob- lems and whether this recognition would activate the kinds of cognitive pro- cesses that are believed to be conducive to SLA. Think-aloud protocols were used as a technique to elicit information about internal processes taking place in the learners. The learners in this study wrote an article for a newspaper about environmental problems and were asked to think aloud as they wrote their articles. From the think-aloud protocols, Swain and Lapkin abstracted what they called “language-related episodes,” in which the learners either spoke about a language problem they encountered while writing and solved it correctly or incorrectly, or simply solved it correctly or incorrectly without having explicitly identified it as a problem. The results revealed that about 40% of these language-related episodes consisted of the students’ paying at- tention to the grammatical form, which suggests that these learners do recog- nize problems in their IL during production. Swain and Lapkin (1995) also found that recognition of problems often triggers cognitive processes that have been implicated in second language learning: processes that either gen- erate linguistic knowledge that is new for the learners or consolidate their ex- isting knowledge. Based on these findings, Swain and Lapkin argue that output not only leads to noticing of the gaps in IL knowledge but also facilitates lan- guage acquisition by triggering various internal processes conducive to SLA. Swain and Lapkin’s (1995) study addressed whether output leads to recog- nition of problems in IL and whether certain internal processes that are be- Effects of Output on Noticing and SLA 425 lieved to be relevant to language acquisition can be triggered. However, this study did not address the question of whether the awareness of problems dur- ing production can prompt the learner to seek out subsequent input with more focused attention. That is, the question remains: Does learners’ recogni- tion of linguistic problems prompt them to notice relevant features if input is subsequently provided to them? The current study attempts to investigate this question by providing learners with opportunities for output, which are then followed by opportunities to receive relevant input, to see whether they would notice and learn the targeted feature in the input. The general research questions to be investigated in this paper are the fol- lowing: 1. Does learner output promote the noticing of linguistic form when relevant input is subsequently provided? 2. Does output result in the acquisition of the linguistic form? RESEARCH HYPOTHESES The Output Hypothesis generates the following predictions concerning notic- ing and SLA. The first hypothesis is derived from the first research question and is concerned with the noticing issue, and the second and third hypothe- ses originate in the second research question and are concerned with the ac- quisition issue. Hypothesis 1 The experimental group, which is allowed to produce output, will show more noticing of the target grammatical form contained in the input than the con- trol group, which does not produce output but instead answers true/false comprehension questions. Hypothesis 2 The experimental group will indicate immediate incorporation of the target form in their output during the treatment tasks after being exposed to the in- put containing the target form. Hypothesis 3 The experimental group will show greater gains than the control group in the accuracy of the use of the target form on posttests. 426 Shinichi Izumi et al. METHOD Participants Participants for the study (N = 22) came from two college-level academic writ- ing classes in a community college ESL department. The participants came from the second level of three writing courses in an ESL program. Students were placed into the course by a departmentally administered placement test or achieved placement by passing the previous writing course. Most of the students in this program plan to enter a two- or four-year degree program after completing their required ESL courses. Participants in the study came from the following first language backgrounds: Spanish (8), Vietnamese (6), Amharic (2), Chinese (2), Turkish (1), Serbo-Croatian (1), Tagalog (1), and So- mali (1). Ages of the participants ranged from 18 to 55. They varied in their amount of formal education, time in the United States (1.7–8.5 years), and in number of EFL or ESL courses taken. All but two spoke their native language at home. After the administration of the pretest, participants were assigned to an experimental group (EG, n = 11) and a control group (CG, n = 11) using a stratified random assignment procedure with the pretest results serving as the basis of stratification. Target Form The past hypothetical conditional in English was the target form for this study (e.g., If Ann had traveled to Spain in ’92, she would have seen the Olympics). Celce-Murcia and Larsen-Freeman (1983) noted that conditional sentences in general, and hypothetical or counterfactual conditionals (imaginative condi- tionals) in particular, cause problems for many ESL learners. Conditional sen- tences consist of two clauses: an if-clause and a main clause. Therefore, the structure of such sentences is complex. Mastery of this structure requires a good grasp of the English tense system and the modal auxiliaries. The pretest, administered prior to the treatment, showed that participants in the study had gaps in their ability to use the target form. Although their attempted use of the structure suggested that this form was not completely new to them, all of them still had problems using the form accurately. Some examples of the IL sentences produced by the participants are as follows: (1) If Kevin got up early in the morning, he would eat breakfast. (2) If Jack travels to China, he would saw the Great Wall. (3) If Kevin have caught the bus, he must have been on time at school. Such emergence of the target form may indicate a certain readiness to learn the form, because many of the participants already attempt to use it in a meaningful way. Effects of Output on Noticing and SLA 427 Table 1. Experimental sequence Day Activity Time 1 Pretest 40 min 2 Treatment phase 1, Task A 30 min Paragraph about Christopher Reeve 3 Treatment phase 1, Task B 30 min Paragraph about Clinton’s welfare bill 4 1st posttest 40 min 5 Treatment phase 2, Task A 1 hr Essay on job offers 6 Treatment phase 2, Task B 1 hr Essay on scholarships 7 2nd posttest 40 min Experimental Design The design of the study was a true experiment with pre- and posttests and a control group. As an overview, Table 1 shows that the experimental sequence was carried out over a period of approximately 1 month, totaling approxi- mately 5 hours. As can be seen from Table 1, the treatment consisted of two phases, with two tasks in each phase. Each treatment was followed by a posttest. In an at- tempt to minimize the test effects, the first phase of the treatment began a full week after the pretest, and the second phase of the treatment began a full week after the first posttest. Treatment All the materials used in the study were designed specifically for adult, immi- grant, ESL learners such as those in this specific ESL program. The written modality was used throughout the experiment for both input (i.e., reading pas- sages) and output (i.e., writing tasks). Piloting of all the materials in the study was carried out with 65 participants from other classes of the same level in the same program. Both phases of the treatment were intended to give the EG participants the opportunity to produce output, hence providing them with the chance to no- tice a gap in their ILs (see Figure 1). Exposure to targetlike input was intended to allow participants to compare their ILs to the target forms in the input. The CG did not produce, but instead answered true/false comprehension ques- tions. In order to assess noticing of the target form, participants were required to underline the passage each time they were provided with the input. The direc- tions given to the EG were to “underline the word, words, or parts of words that you feel are particularly necessary for your subsequent (re)production.”1 The CG was also required to underline their passage for comprehension (i.e., 428 Shinichi Izumi et al. Figure 1. Instructional procedures. to answer questions about the passage). Before participants carried out the tasks, the underlining portion of the activity was modeled for both groups. The researchers, using a passage that did not contain the target form, showed participants examples of underlining, to illustrate the options of underlining the “word, words, and parts of words” of the passage and to stress the impor- tance of precise underlines. This was done to enhance participants’ familiarity with the underlining procedure and the preciseness of this measurement of noticing. The entire process was role-played with one researcher giving the directions and another researcher carrying out the task. This was to help the participants realize that they would not be able to look at their underlining passage while reproducing or to check their first production while doing the second. It was necessary to clarify these two points to collect the most accu- rate data possible, as was discovered when the tasks were piloted with a simi- lar class. Both the EG and the CG were informed in