语言输出与外语学习(txt+pdf+epub+mobi电子书下载)

作者：刘春燕

出版社：世界图书出版广东有限公司

格式: AZW3, DOCX, EPUB, MOBI, PDF, TXT

语言输出与外语学习试读：

前言

人们对语言输出的作用一直存在争议：输出是习得的结果还是习得的原因？语言输出过程与语言学习过程是否相关？语言学习的认知理论对注意、记忆、输入、表征系统和技能习得等进行了大量的研究，使得二语习得的认知研究既具解释性又具操作性。但该理论却认为输出是二语习得的结果表现，或最多只是提高学习者语言的流利度。另一方面，Swain（1985, 1993, 1995, 1998）提出的“输出假设”认为，语言产出在二语习得中起着多种积极的作用。但输出是否直接导致中介语的发展、输出时中介语变化的心理语言学机制是什么等问题仍旧没有明确的结论。本项目是一项关于语言输出在二语习得中的作用的研究，该研究旨在探索语言输出在外语词汇学习、言语行为和认知发展中的作用，并在实证研究的基础上建立起语言输出在二语习得中的作用的信息加工理论模型。

本书共分八章。第一章简要介绍了二语习得的输出假设及该假设未能解决的问题。第二章对二语习得领域的主要理论作了简要的回顾并揭示语言输出在这些理论中的地位。第三章为本研究理论框架，在语言输出及其认知过程在二语习得中的作用的相关文献的基础上，提出本研究的概念框架，即语言输出的理论模型。第四章汇报了实验1的研究设计，探讨了语言输出任务在习得生词中的作用。第五章汇报了实验2的研究目的、研究问题及假设，探讨了输出任务在提高整体言语行为方面的作用。第六章介绍了出声思维方法及作文的文本分析方法以揭示语言输出过程中的认知过程和认知技能的发展，并对出声思维数据和作文文本进行分析与讨论。第七章调查了现有英语课程体系中输出任务的地位。第八章为本研究结论和对语言教学启示，提出基于输出的语言教学模式和输出干预型的教学。

具体说来，本研究包括以下研究问题：

1）输入任务组与产出任务组的生词习得数量是否存在显著差异？输出任务中哪些条件有助于生词的习得？

2）与理解任务组相比，产出任务组是否在流利度、准确度和复杂度等言语行为方面有更好的表现？

3）学习者产出二语时是否介入精致化过程/重构过程？换句话说，二语的产出过程是否引起学习者习得二语所必需的认知过程？

4）输出任务组是否比输入任务组表现出更好的认知技能特点？

5）输出任务在现有课程体系中的地位如何？课程实施过程中是否对语言输出予以充分的注意？

在研究设计和方法上，本文采用定量研究与定性研究相结合的方法，系统全面地考察输出的结果与过程等各个方面。定量研究包括两个实验设计：实验1对比4组（共181名被试）英语专业大一学生在不同条件下（输入、理解任务、无意注意产出任务和有意注意的产出任务）习得生词的效果，从而调查输出任务在习得生词中的作用；实验2考察两组（25名被试）英语专业的学生在一个月的理解任务和产出任务处理后中介语的变化情况。两个实验的前测和后测的结果构成了定量研究的实证数据。定性研究部分包括出声思维、访谈和作文文本分析等数据。出声思维数据来自4名被试出声地完成重写任务（text reconstruction task）和回顾性访谈。本研究根据被试重写时所采用的交际策略和问题解决机制，将出声思维数据分为句法加工、词汇加工和策略转换等三类语言加工过程，并计算每一类过程的频数和比例。作文文本分析是对50篇作文（实验2中的25名被试的前测和后测作文）的思维技能变化特点的分析。本研究根据思维特点的六种表现形式（即论点表述清楚、用事实和推理支持观点、提及反面观点、作文不是包罗万象或百科全书式的写作、内容不是资料堆积式的毫无论点、句子和段落间的连贯等）对作文文本进行对比分析。

本研究有如下研究结果与发现：

首先，理论模型中的主要部分得到了验证。1）实验1的结果表明输入组与两个产出组的生词习得数量存在显著差异（p＜.000）。这表明产出并非是对已有知识的简单提取，而是对新知识和已有知识的深加工。输出过程迫使学习者在具体的情景下对目标语形式精细化加工和使用，并通过语言使用而在习得新的语言形式时起着独特的作用。2）实验2的结果表明，产出任务组的言语行为总体好于理解任务组，特别在流利度和精确度方面。产出任务组的前测与后测的比较结果表明，产出任务对言语行为的三个测量指标（流利度、准确度和复杂度）均有显著效果。3）对出声思维的数据分析表明，输出过程使学习者进行相应的语义（词汇）加工、句法加工和策略转换，从而使学习者对现有知识进行重构。4）与理解任务组相比，产出任务组的后测作文表现出更好的思维特点，特别在作文的连贯性方面。产出组的前测均值与后测均值的对比结果也表明产出任务对认知发展有显著的效果（p＜.01）。

第二，输出与其他认知系统的关系也得到进一步的验证。1）实验2中产出任务组的理解测试与理解任务组的测试结果没有差异，这说明产出能“触发”学习者深入加工语言输入，以便于更好地输出。2）实验1中无意注意产出任务组和有意注意的产出任务组所习得的生词数没有显著差异。这说明产出任务能使学习者主动地分配注意资源和控制自己的学习，亦即有元认知的参与与“监控”。3）出声思维数据中，两名被试没有实现有关宾语从句的知识表征系统重构，但她们已感觉到自己产出的语言与新遇到的结构间的差距。被试预期水平与现实之间的这一差距成为发展中介语的动力，即输出任务促使学习者“注意到差距”，并促使被试更加关注后续的相关输入，或急切需要教师的干预和帮助，而这正是教师教学的最佳干预点。

第三，对输出和语言使用在我国现有课程体系中的地位和作用的调查表明，虽然语言使用和输出在各级英语大纲中都有明确要求，但是在课程实施过程中却没有对语言使用和输出予以充分的注意和重视，特别在课堂教学和大规模考试中语言输出严重不足。究其原因，传统的中国文化与现有教学理念使得人们重视输入性的理解教学而忽视复用型的输出教学。

本书是在作者的博士论文的基础上改写而成的，它具有如下几个特点：

1. 全面、详尽、深入地总结了二语习得研究领域对语言输出研究的成果和不足，使人们对“输出假设”及语言输出在二语习得中的作用有比较清晰的了解。无论是对语言输出的理论研究进程，还是对语言输出的实证研究，本书都作了详实的介绍。

2. 介绍了作者所做的关于语言输出的定量研究与定性研究，系统全面地考察了语言输出的过程与结果等各个方面。定量研究包括两个实验设计，其前测和后测的结果构成了定量研究的实证数据。定性研究部分包括出声思维、访谈和作文文本分析等数据。此外，为了使本研究更具实践意义，作者考察了输出在我国现有课程体系中的地位和作用。我们考察了语言产出和语言使用在各级大纲的要求、产出任务在现行大学教材设计中的比例、产出活动在课堂教学中的比重及其原因、产出活动在各级大规模考试中的比例等，为进一步研究语言输出和语言使用能力的培养提供前期数据。

3. 作者提出了以语言输出为中心的信息加工二语习得理论模型。该模型将输出及输出的研究置于信息加工学习理论的框架下，并建立在作者的实证研究基础之上。该理论模型既吸收了认知主义的二语习得观中强调语言学习过程主要是大脑的认知加工过程的观点，又吸收了社会文化论认为语言使用过程也是语言习得过程的观点，突出语言输出在二语习得中处于核心地位，认为它既是习得的结果，更是习得的原因。从理论上讲，输出在二语习得中的作用在认知理论中多半被否认和被忽视。以语言输出为中心的二语习得理论模型的提出使人们重新认识输出在习得中起着举足轻重的作用。

4. 本书提出的基于输出的语言教学模式和输出干预型的教学对改进现有的英语课程设计和实施提出了切实可行的建议，对提高外语学习效率、改变外语学习费时低效现象起到积极作用。

本书的出版得到江西师范大学国际教育学院的科研基金的部分资助和世界图书出版公司（广州公司）的大力支持。作者的导师上海外国语大学原校长戴炜栋教授对本研究进行了精心指导和极大的勉励，并欣然作序。世界图书出版公司（广州公司）的总编辑萧宿荣博士、江西师范大学国际教育学院李行亮院长和黄焰烘副书记对本书予以热情的支持和帮助，在此一并致以衷心的感谢。

本书有助于我国从事二语习得的研究人员、外语教师、高校外语专业的研究生了解国内外关于语言输出的研究成果，提高二语习得研究水平和推动我国外语教学改革。

Acknowledgments

I'm most indebted to my supervisor, Prof. Dai Weidong, who instructed and supported me through the days of planning and writing. He was kind and generous to spend his time discussing my research work with me whenever and wherever I needed assistance. Without his guidance and encouragement, this dissertation would never have been finished. I am grateful to Professor P. Skehan, now at the Chinese University of Hongkong, who kindly commented on and discussed with me about Experiment 2 in Chapter 5. In addition, I am grateful to Professor R. Ellis, who made valuable suggestions to me at the Second International Conference on Foreign Language Teaching Methodology in China. I am grateful to Dr. Cai Jinting, who generously offered me help in statistics, i.e., the special z-test and the z-test program which he and his friends developed by themselves.

My hearty thanks also go to the following people who helped me in various ways in my completion of this piece of work:

To my friends, Liu Fulan and Wu Limin, who collected related materials for me from Melbourn University in Australia and Guangdong University of Foreign Studies, and the teachers in Jiangxi Normal University who kindly cooperated with the experiments.

To many colleagues and friends, Prof. Liu Shulin, Prof. Rao Zhenhui, Dr. Zhou Dajun, Dr. Yang Xianju, Dr. Wang Shunyu, Dr. Tang Xiongying, Dr. Zhang Yan, Dr. Chen Kefang, Dr. Cai Junmei, Dr. Ge Xianru, and Dr. Zhou Caiqing, who have assisted through discussion, argument, and sometimes challenge.

To some of my graduate students in Jiangxi Normal University who helped me collect the research data, and to those cooperative students who have participated in this study.

And lastly, I am particularly indebted to my husband, Dr. Zhong Zhixian, for his unfailing support and encouragement during these years; to my dear daughter, Zhong Chuanxiuxing, for her understanding and support of my academic work; to my parents and other family members for their active roles in minimizing the repercussions of my academic research on a normal family life.

Chapter 1 Introduction to Output Hypothesis in Second Language Acquisition（SLA）

There are good theoretical and educational reasons to place matters of language output high on the agenda for Second Language Acquisition (SLA) research. As for theoretical motivations, the possible roles that output (talking and writing) might play in second language learning are crucial for those scholars who want to construct a SLA theory within information-processing framework. Skehan (1999) and Gass (1988, 1997, 2001) build information-processing models which explain psycholinguistic processes in language use and language learning. In their models, there are different stages in information-processing: input, central processing, and output. The information-processing models are helpful in separating the different stages concerned, and in providing an organizing framework for more detailed discussion of the functioning of each separate stage. Research to date has provided descriptive evidence of the existence of learning processes stimulated by output (Cumming, 1990; Swain & Lapkin, 1995). However, there is a paucity of research that demonstrates whether these output-oriented processes are facilitative of second language learning.

As concerns educational motivations, curriculum planners, material designers, teachers, and learners all have a vested interest in knowing in which linguistic domains second language (L2) learning might best benefit from output learning modes. Swain was the first to point the insufficiency of the French immersion programs in Canada. More than two decades of research in French immersion classes suggests that immersion students are able to understand much of what they hear and read even at early grade levels. And although they are well able to get their meaning across in their second language, even at intermediate and higher grade levels they often do so with nontargetlike morphology and syntax (Swain, 1984). Swain's research, related to the French proficiency of immersion students, makes clear that an input-rich, communicatively oriented classroom does not provide all that is needed for the development of targetlike proficiency. It also makes clear that teaching grammar lessons out of context, as paradigms to be rehearsed and memorized, is also insufficient. In her observation of immersion classes, Swain found that students were given limited opportunities for extended output where linguistic accuracy was demanded. She wondered whether this in itself was having an effect on the nature of immersion students' proficiency. This motivated her to propose the “output hypothesis” to explain the insufficiency of the immersion program.

1.1 Swain's output hypothesis

The function of output in enhancing fluency seems non-controversial. We know that fluency and accuracy are different dimensions of language performance, and although practice may enhance fluency, it does not necessarily improve accuracy. According to Swain's “output hypothesis”, three functions of output have been proposed that relate more to accuracy than to fluency in second language learning (Swain, 1995). The three functions are: noticing, hypothesis formulating and testing, and conscious reflection.1.1.1　Noticing

First, it is hypothesized that output promotes “noticing”. This is important if there is a basis to the claim that noticing a form in input must occur in order for it to be acquired (Schmidt, 1990, 1992). The sense in which Swain and Lapkin (1995) have used “noticing” coincides with that of Schmidt and Frota (1986), who state that by “noticed”, they mean “a second language learner will begin to acquire the targetlike 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). According to Schmidt (1993), noticing requires focal attention and awareness on the part of the learner. What is noticed is available for verbal report.

There are several levels of noticing. Learners may simply notice a form in the target language due to the frequency or salience of the features themselves, for example (Gass, 1988). Or, as proposed by Schmidt and Frota (1986) in their “noticing the gap principle,” learners may notice not only the target language form itself but also that it is different from their own interlanguage. Or, learners may notice that they cannot say what they want to say precisely in the target language (Swain, 1995). Swain's output hypothesis is that output gives rise to noticing. That is to say, in producing the target language, learners may encounter a linguistic problem leading them to notice what they do not know, or know only partially. In other words, the activity of producing the target language may prompt second language learners to consciously recognize some of their linguistic problems; it may make them aware of something they need to find out about their L2—that is, “noticing the gap”. The important issue here is that “noticing the gap” may trigger cognitive processes that might generate linguistic knowledge that is new for the learner or consolidate the learner's existing knowledge (Swain & Lapkin, 1995).1.1.2　Hypothesis formulation and testing

A second way in which producing language may serve the language learning process is through hypothesis formulation and testing. It has been argued that some errors which appear in learners' written and spoken production reveal hypotheses held by them about how the target language works. To test a hypothesis, learners need to do something, and one way of doing this is to say or write something. Tarone and Liu (1995, pp. 120-121) provide evidence that it is precisely in contexts “where the learner needs to produce output which the current interlanguage system cannot handle…[and so]…pushes the limits of that interlanguage system to make it handle that output” that acquisition is most likely to have occurred.

During the process of negotiating meaning, learners will modify their output in response to such conversational moves as clarification requests or confirmation checks. For example, Pica and her colleagues (1989) found that in response to clarification and confirmation requests, over one-third of the learners' utterances were modified either semantically or morphosyntactically. The fact that learners modify their speech in one third of their utterances suggests that they are testing out only some things and not others. Swain (1995) proposes that it may be that the modified, or reprocessed, output can be considered to represent the leading edge of a learner's interlanugage. This is further confirmed by Swain & Lapkin (1995) and Izumi (2003). Swain & Lapkin illustrate that when learners notice a problem, they conduct an analysis leading to modified output. This may include three general categories of cognitive processes: generating alternatives, assessing those alternatives, and applying the resulting knowledge. Izumi (2003) demonstrates that what goes on between the original output (Output 1) and its reprocessed form (Output 2) is part of the process of second language learning. When facing problems in their production process, learners have several alternative routes to take depending on the given situation at the time. In other words, output processing engages important internal procedures such as grammatical encoding and monitoring, which prompts the learners to interact actively with the external environment to find a solution or to explore their internal resources for possible solutions. Output, thus, serves as a useful means to promote the interaction between learner internal factors and environmental factors, or the interaction within the learners themselves for internal metalinguistic reflection. The processes that intervene between the first output and the modified output are believed to constitute an important part of SLA.

Swain (1995) summarizes that learners may use their output as a way of trying out new language forms and structures as they stretch their interlanguage to meet communicative needs; they may use output just to see what works and what does not. That immediate external feedback may not be facilitative or forthcoming does not negate the value of learners having experimented with their language resources.1.1.3　Conscious reflection

A third function of output is its metalinguistic function, in other words, students talk about language, or “negotiate about form”. It is claimed that “as learners reflect upon their own TL (target language) use, their output serves a metalinguistic function, enabling them to control and internalize linguistic knowledge” (Swain, 1995:126). In other words, output processes enable learners not only to reveal their hypotheses, but also to reflect on them using language. It is this “level” of output that represents its metalinguistic function of using language to reflect on language, allowing learners to control and internalize it. In consciousness-raising tasks (Ellis, R. 2003), for example, learners communicate about language, in the context of trying to produce something they want to say in the target language. Learners negotiate meaning, but the content of that negotiation is language form, and its relation to the meaning they are trying to express; they produce language and then reflect upon it. They use language to “negotiate about form.” Reflection on language may deepen the learners' awareness of forms, rules, and form-function relationships if the context of production is communicative in nature.

Swain's working assumption is that conscious reflection about language (or metatalk) is a surfacing of language used in problem solving; that is, it is language used for cognitive purposes. In metatalk, we are able to observe learners' working hypotheses as they struggle toward solving mathematical problems, scientific problems, or as we are concerned with in SLA, linguistic problems. If this is the case, then much of what is observed in metatalk when learners are faced with a challenging language production task and are encouraged to talk about the problems they encounter in doing the task should help us to understand language in progress. In other words, in metatalk, noticing, hypothesis formulation and testing (cognitive problem solving), and other learning processes may be made available for inspection. They are available for inspection by researchers, teachers, and possibly most important, for students themselves as they engage in second language learning. Thus, by encouraging metatalk among second and foreign language students, we may be helping students to make use of second language acquisition processes. That is, metatalk may be one pedagogical means by which we can ensure that other language acquisition processes operate.1.1.4　Some caveats and comments

When proposing Output Hypothesis, Swain (1995) also made several comments to prevent misunderstanding of Output Hypothesis. The first is that no claim is being made that any or all of these functions operate whenever learners produce the target language. It will be one of the tasks of future research to determine under what conditions they do operate.

The second comment is that output-based studies are necessary for language pedagogy and SLA. Research has shown that a communicatively-oriented input-rich environment does not provide all the necessary conditions for second language acquisition, and that a focus on form within these communicative settings can significantly enhance performance. The output-based studies like collective scaffolding provide important evidence for the usefulness of collaborative tasks that lead learners to reflect on their own language production as they attempt to create meaning.

Her third comment relates to the tests we use in measuring learning. She thought it would seem crucial if we are to measure the learning which occurs as a result of task involvement, and that we must consider tailor-making our tests to actual task performance.

The fourth comment relates to the data we use to inform ourselves of learners' cognitive processes. She thought that the unit of analysis of language learning and its associated processes may more profitably be the dialogue, not input or output alone.

To add to the list, Izumi (2003) also made it clear that Output Hypothesis in no way negates the importance of input or input comprehension. The intention is to complement and reinforce, rather than replace, input-based approaches to language acquisition so that learners will go beyond what is minimally required for overall comprehension of a message.

In revealing the underlying psycholinguistic rationale of the Output Hypothesis, Swain (1985, 1995) and Izumi (2003) argue that production makes the learner from “semantic processing” prevalent in comprehension to more “syntactic processing” that is necessary for second language development. Output pushes learners to process language more deeply (with more mental effort) than does input. By drawing on Levelt's (1993) speech production model, Izumi argues that the processes of grammatical encoding during production and monitoring to check the matching of the communicative intention and the output enable language learners to assess the possibilities and limitations of their interlanguage capability. This may, under certain conditions, serve as an internal priming device for consciousness raising for the learners, which in turn creates an optimal condition for language learning to take place.1.1.5　More roles for output

In addition to the three functions of Output Hypothesis, Skehan (1999) added two more functions of output:

1）to develop discourse skills

Skehan emphasizes the discourse skills in conversations, such as discourse management, turn-taking skills, and a range of similar capacities which underlie the negotiation of meaning in ongoing discourse. He argues that these skills can only be achieved by actually participating in discourse. Likewise, discourse skills in writing should also be achieved in writing practice. These include cohesion and coherence, the macrostructure of the composition, the logic of the content, the knowledge of genre, etc. “If meaning-making is a jointly collaborative activity, then we cannot read about these skills, or even acquire them passively, but instead have to take part in discourse and realize how our resources are put to work to build conversations and negotiate meaning. Extensive speaking practice is therefore unavoidable.” (p.18)

2）to develop a personal voice

It seems inevitable that if one wants to say things that are important, one must have, during language learning, the opportunity to steer conversations along routes of interest to the speaker, and to find ways of expressing individual meanings. A role for output here seems unavoidable. This function has been neglected in language teaching in China. Learners in our country are dependent on the sorts of meanings that he or she has been exposed to, and will not able to exert an influence on conversational topics. Thus, learners become passive and dependent because they seldom to have the chance to express themselves. A learner who is completely dependent on what others say, is unlikely to be able to develop a personal manner of speaking.1.1.6　Unresolved questions

The Output Hypothesis is not without criticism. The most important problem is that the studies on Output Hypothesis did not directly address the issue of whether output leads to acquisition or not. As Shehadeh (2002) points out, existing research was mostly descriptive in nature, focusing primarily on occurrence per se rather than acquisition or whether and how interaction/output can be a source of competence in the L2. Much of the empirical research confined itself to examining the different learner and contextual factors that affect learner's interaction (e.g., gender difference, task type, and discourse variables) rather than investigating whether and how interaction/output can help with L2 learning.

R. Ellis (2003) also points out that, although Swain's Output Hypothesis is persuasive on the face of it, it fails to provide a convincing explanation for how production leads to acquisition. “Swain (1995) falls back on the metaphor of “stretching interlanguage” but never explains what this means… In short, while the importance of production for acquisition has been clearly established, it would seem to play only a limited causal role, motivating learners to attend to input, but not contributing directly to the processing needed for acquisition to take place.” (p. 114)

Summarizing the opposing views about the role of output in SLA, here are some unresolved questions as to output and its relationship with other cognitive systems.

1）Is output the process of acquisition or the result of acquisition?

Output has been viewed as the result of acquisition for a long time. In the first language acquisition, Cazden (1968) linked the point of acquisition with output. He defined the point of acquisition of several noun and verb inflections as “the first speech sample of three such that in all three the inflection is supplied in at least 90 percent of the contexts in which it is clearly required (p 435)”. With only a slight modification, the point of acquisition in SLA was adapted and defined by Hakuta (1974) as the “first of the three consecutive two-week samples in which the morpheme is supplied in over 90% of obligatory contexts” (p. 137). Other researchers have tried to focus on the emergence of structures, rather than on their mastery, and defined acquisition of a form as the first appearance of that form in the learner's language. All these definitions of acquisition draw on learners' output. In other words, output is viewed as the result of acquisition.

Early theories of second language acquisition, such as Krashen's Monitor Model, emphasizes “comprehensible input” as the trigger of language acquisition. Acquisition is caused by understanding the input to which the learner is exposed. Internal factors are given little emphasis. The importance of output is de-emphasized. Krashen (1985), when emphasizing the crucial role of input, claimed that “speaking is a result of acquisition, not its cause” (p2).

In contrast, some cognitive psychologists/SLA researchers believe output can also impact the underlying grammatical system and lead to second language acquisition. Hulstijn (1990) argues that output involves “the establishment of new procedures which reorganize a body of facts and rules previously acquired”. McLaughlin (1987) has written:But there is more to learning a complex cognitive skill than automatizing subskills. The learner needs to impose organization and to structure the information that has been acquired. As more learning (in 1990, he says “practice”) occurs, internalized, cognitive representations change and are restructured. This restructuring process involves operations that are different from, but complementary to, those involved in gaining automaticity. (p.136)

An example of restructuring is the development of organizational schemata, higher-order abstract representations of complex sub-elements. Another example of abstract patterns that might result from the organizing and structuring of information is the development of prototypes.

Modern psychology also emphasizes the role of active output and production in language learning. Constructivism (see Liu and Liu, 2004), which is a new development of cognitivism, argues that knowledge is actively constructed by the learner, not passively received from the environment, and that coming to know is a process of adaptation based on and constantly modified by a learner's experience of the world. Connectionism, a newer modeling approach from psychology, views learning as a consequence of repeated neural network activation that results in stronger, and therefore more easily activated, connections. In connectionist model, the process of output and production is the process of learning.

2）How is output process related to learning process? Does output relate to other subsystems of information-processing system such as noticing, memory, and input, central processing, etc.?

For researchers working in the generative framework (e.g. Gregg, 1996; White, 1996), the learning mechanisms are structurally denied, and include parameter-(re)setting, triggering, or the Subset principle. How these mechanisms actually work in real-time processing remains largely a matter of conjecture. Output is not needed for language acquisition, or at best is out of synchronization with the natural development of grammatical competence. What's more, output pushes learners to speak before they are ready, which might lead to negative affect and misrepresentation of the grammatical rule. In a word, output does not directly affect the competence system itself.

For researchers working in cognitive framework, although early cognitive theories largely ignored developmental issues, there is a growing recognition that the understanding of processing and learning is closely intertwined, if not identical. The relationship between processing and learning mechanisms is more direct in the connectionist approach, where they represent two sides of the same representational coin. The knowledge that subserves sentence processing is stored in a distributed network of activations that are graded, thus varying in strength as a function of the individual's experience. Processing outcomes are the result of the interaction of multiple sources of information (lexical, syntactic, frequency, context) that are a direct reflection of the individual's knowledge at that point. That knowledge, expressed in levels of activation, changes in principle every time the processing mechanism is engaged. In this view processing is learning. The learning/processing mechanism responsible for the development of this knowledge is the accretion of patterns of activation through associative learning (Elman, 1990; MacDonald, 1997).

Some scholars view communication strategies as the mechanism of language learning during output and production tasks. Tarone (1988) relates variation theory to second language acquisition through communication strategies. She argues that a learner's use of communication strategies can function to stretch an interlanguage (IL) system beyond its current limits, resulting in free variation as the learner tests new hypotheses in the search for an appropriate word or structure. She bases her argument on data presented by Liu (1991) who studied a young Chinese child, Bob, over two years as he learned English (see Tarone and Liu, 1995).

It is these problems that motivate so many second language researchers and teachers, including the author, to explore the Output Hypothesis and beyond. We do not have complete solutions yet, and, as such, the key issue is to decide how we can make progress, and add to what we know about output, rather than dismiss it because we do not yet have the whole story.

1.2 Structure of this book

The organization of the book is as follows. The first 3 chapters of the book introduce fundamental concepts and theoretical issues on Output Hypothesis and the role of output, followed by Chapters 4-7, empirical studies of the role of output. The last two chapters are applications to instruction and conclusions which attempts to provide a more unified picture of information-processing view of second language learning.

Looking at the outline in more detail, Chapter 1 examines Output Hypothesis in language learning. It explores the three functions proposed by Swain (1995), and more roles proposed Skehan (1999). It moves on to look at the attacks that were mounted on studies of Output Hypothesis, and the unresolved questions. Hence, Output Hypothesis is limited in its explanatory power, and more complex approaches are necessary to do justice to the relationship between output and acquisition in second language learning.

Chapter 2 reviews some of the major SLA theories found in the literature and the role of output in these theories. These theories include Input Hypothesis, Vygotsky's Zone of Proximal Development, Piaget's View of Language Acquisition, UG-based SLA research, Skehan's Cognitive Approach, Gass's Integrated Model, Connetionism, and Sociocultural Theory. It suggests that earlier claims, by Krashen and UG-based SLA researchers, for example, emphasize the crucial role of input and the internal factors in acquiring second language. In these theories, output is the result of acquisition, not the cause of acquisition, and little emphasis is given to output. Later theories like Connectionism and sociocultrual theory hold that relationship between second language use (mainly output) and second language development is more direct, and that language processing is learning.

Chapter 3 sketches out a conceptual framework depicting the roles of output in language acquisition and proposes the information-processing framework of SLA. It first reviews empirical studies on the role of output in the literature, then examines the underlying psycholinguistic rationale of comprehension and production. Following researchers such as Vygotsky and Piaget, the chapter argues for the central importance of output in promoting language performance and cognitive development of the learner. The chapter concludes with a conceptual framework of the role of output in second language acquisition.

The following 4 chapters in the book provide empirical studies of how output tasks contribute to different aspects of language acquisition and learner development. Chapter 4 introduces an experiment (Experiment 1) which is designed to examine whether output tasks contribute to the acquisition of new words and what conditions lead to acquisition of new words in output tasks. Four groups of subjects (the control group, the comprehension group, the production group, and focused-attention group) were compared to see how different tasks and different conditions influence the acquisition of new words. Drawing on Laufer and Hulstjin's (2001) involvement load in vocabulary acquisition, we showed how output tasks could promote word acquisition through elaboration and language use.

Chapter 5 reports the results and discussion of Experiment 2, which is designed to explore how output tasks (text reconstruction tasks) differ from input tasks (reading comprehension tasks) in promoting language performance on the part of the learner. Language performance is analyzed from three aspects: fluency, accuracy and complexity. The experiment shows how output tasks have significant effect on all the three measures and on learner's interlanguage development

Chapter 6 deals with the cognitive processes and cognitive development in output. Drawing on Vygotsky's and Piaget's theory, the study tries to investigate the underlying cognitive processes during output through the think-aloud techniques. Also, text analysis of the pre-test and posttest compositions shows how the cognitive skills and abstract thinking are involved in writing compositions, and how cognitive development occurs in language output.

Chapter 7 surveys the application of output tasks in the curricula of different levels in China. It first examines the requirements of productive skills and knowledge in the syllabuses of different levels in China. Then it reports the implementation of the curriculum in the textbooks, in the classroom and in the examinations. Finally it provides suggestions and implications for application of output tasks in English curriculum design and implementation.

Drawing on Chapter 7, which focuses on the neglect of meaningful output practice in FLT in China, Chapter 8 tries to take these earlier discussions of the role of output into account, and explores what implications they have for instruction. It first summarizes the major findings of the study, and moves on to what can be done in language teaching and learning. These include Output-oriented instruction, principles of Output-oriented instruction in language teaching, Output-oriented model in language teaching, and the implementation of task-basked language teaching. Future directions for inquiry are made both on the macro-level research of theoretical aspect and micro-level research of pedagogical aspect.

Chapter 2 Theories of SLA and the Role of Output

This chapter reviews some of the major SLA theories found in the literature and the role of output in these theories. A great many theories regarding language development in human beings have been proposed in the past and still being proposed in the present time. Such theories have generally arisen out of major disciplines such as psychology and linguistics. Psychological and linguistic thinking have profoundly influenced one another and the outcome of language acquisition theories alike.

The recent development of SLA research embodies its multidisciplinary characteristics. Although most researchers conceptualize SLA as an independent field with its own research agenda and with its own research focus, SLA has strong ties to other disciplines such as psycholinguistics, theoretical linguistics, sociolinguistics, cognitive linguistics, communication studies, psychology, and education. And SLA can also be thought of as a subdiscipline of one source discipline or another. Among these source disciplines, three areas exert strong influence on SLA: linguistics, psychology and social constructivism. One feature all these areas share is a concern with the learning problem. That is, how can learners acquire the complexities of a second human language? (Gass & Selinker, 2001) In this chapter, we generally focus on the influence on SLA from these disciplines, with the emphasis on the recent development of SLA research. These theories include:

Linguistics-based Research:

-The Monitor Theory

-UG-based SLA research

Cognitive Perspective:

-Piaget's View of Language Acquisition

-Skehan's Cognitive Approach

-Gass's Integrated Model

-Connetionism

Sociocultural Perspective:

-Vygotsky's Zone of Proximal Development

-Sociocultural Theory

2.1 The Monitor Theory

One of the best known and most influential theories of SLA in the 1970s and early 1980s was Krashen's Input Hypothesis or Monitor Theory (MT) (1985). MT was also one of the first “theories” developed specifically to explain SLA. There are five basic hypotheses in this model: 1) The Acquisition-Learning Hypothesis, 2) The Natural Order Hypothesis, 3) The Monitor Hypothesis, 4) The Input Hypothesis, and 5) The Affective Filter Hypothesis.

1）The Acquisition-Learning Hypothesis

Krashen, in his theory of second language acquisition, suggested that adults have two different ways of developing competence in second languages: Acquisition and learning. “Acquisition” is a subconscious process identical in all important ways to the process children utilize in acquiring their first language, while “learning” is a conscious process that results in “knowing about” language (Krashen, 1985:1). But what Acquisition/Learning Distinction Hypothesis predicts is that learning the grammar rules of a foreign/second language does not result in subconscious acquisition. In other words, what you consciously learn does not necessarily become subconsciously acquired through conscious practice, grammar exercises and the like. Krashen formulates this idea in his well-known statement that “learning does not became acquisition”. It is at this point where Krashen receives major criticism.

2) The Natural Order Hypothesis

According to this hypothesis, the acquisition of the rules of language proceeds in a predictable order, some rules tending to come early and others late. The order does not appear to be determined solely by formal simplicity and there is evidence that it is independent of the order in which rules are taught in language classes. The “natural order” was determined by a synthesis of the results of the morpheme order studies and is a result of the acquired system, without interference from the learned system.

3）The Monitor Hypothesis

As mentioned before, adult second language learners have two means for internalizing the target language: acquisition and learning. The “monitor” is an aspect of this second process. It edits and makes alterations or corrections as they are consciously perceived. Krashen (1985) believes that “fluency” in second language performance is due to “what we have acquired”, not “what we have learned”: Adults should do as much acquiring as possible for the purpose of achieving communicative fluency. The acquired system was typically the only knowledge source speakers could use in real-time communication, when they were attending to meaning, not form; the learned system served only as a planner and editor with which to inspect, or monitor, the output of the acquired system. Therefore, the monitor should have only a minor role in the process of gaining communicative competence. Similarly, Krashen suggests three conditions for its use: (1) there must be enough time; (2) the focus must be on form and not on meaning; (3) the learner must know the rule.

4）The Input Hypothesis

The central claim of Monitor Model is the Input Hypothesis. Krashen (1981) regarded it as the single most important concept in second language acquisition at the time, in that it attempts to answer the critical question of how we acquire language. It maintains that a second language is acquired through processing comprehensible input, i.e. language that is heard or read and understood. Krashen claims that people acquire language best by understanding input that is a little beyond their present level of competence. Consequently, Krashen believes that comprehensible input (that is, i + 1) should be provided. If input is understood and there is enough of it, the necessary grammar is automatically provided. Speaking is a result of acquisition and not its cause. The ability to communicate in a second language cannot be taught directly but “emerges” on its own as a result of building competence via comprehensible input.

However, Krashen says nothing about how the input is processed and internalized. This is because Krashen assumes a heavy innate endowment to handle acquisition. Krashen (1985, pp. 2-3) writes:Input is the essential environmental ingredient. The acquirer does not simply acquire what he hears—there is a significant contribution of the internal language processor (Chomsky's language Acquisition Device: LAD). Not all the input the acquirer hears is processed for acquisition, and the LAD itself generates possible rules according to innate procedures.

He suggests that what he considers to be the extensive evidence for the Input Hypothesis “supports Chomsky's position, and extends it to second-language acquisition”. Like Chomsky, Krashen further assumes that the endowment is language-specific.

Krashen's Input Hypothesis initially stimulated a good deal of data-based research, and forced some fresh thinking in language teaching circles. McLaughlin's cognitive approach to second language acquisition and Canal & Swain's Output Hypothesis are somehow based on criticism of Krashen's Input Hypothesis.

5）The Affective Filter Hypothesis

The learner's emotional state, according to Krashen, plays an important role in acquiring a second language. According to the Affective Filter Hypothesis, comprehensible input may not be utilized by second-language acquirers if there is a “mental block” that prevents them from fully profiting from it (Krashen, 1985). The affective filter acts as a barrier to acquisition: if the filter is “down”, the input reaches the LAD and becomes acquired competence; if the filter is “up”, the input is blocked and does not reach the LAD. These affective factors are self-confidence, motivation and anxiety state. Lack of motivation, low self-esteem, debilitating anxiety, and so on, can combine to raise the filter, to form a “mental block”.

In conclusion, Krashen adopts a nativist view in explaining second language acquisition. Like other nativist approach to language acquisition, MT suffers from the methodological problems: its five hypotheses are untestable. It purports to account for extensive empirical data and to have broad implications for language instruction, but its claims are unwarranted, and the Monitor Model fails to meet the canons of theoretical sufficiency (McLaughlin, 1987).

The Monitor Theory with its five Hypotheses makes a strong claim—that acquisition is caused by understanding the input to which the learner is exposed. Internal cognitive factors are given little emphasis. The importance of output is also de-emphasized, and understanding a new language is given far greater stress than speaking it.

2.2 UG-based research

Universal Grammar (UG) theory does not concern itself with second language acquisition. “Acquisition part” is of secondary importance in this theory. The application of the theory to SLA has come about through recent work of a number of second language researchers. Four considerations motivate interest on the part of SLA researchers in the Chomskyan position (White, 1985, reported in McLaughlin, 1987):

1) The need for a sufficiently sophisticated linguistic theory to describe the complex structural characteristics of interlanguages. Universal Grammar provides a sophisticated and detailed linguistic theory to account for second language phenomena.

2) The growing realization that second language learners face a “projection problem”—that is, that they, like first language learners, have to work out a complex grammar on the basis of deficient data. The learner's grammatical knowledge cannot be explained by the input data alone.

3) The development of principle and parameter theory within generative grammar, which allows for a more precise investigation of language variation—including variation between native and target languages. Universal Grammar is seen to set the limits within which human languages can vary.

4) The realization that a rigid critical period hypothesis is open to question, so that one does not have to assume that adult language learning is of a totally different or disadvantaged nature when compared to child acquisition.

The theory underlying UG assumes that language consists of a set of abstract principles that characterize core grammars of all natural languages. In addition to principles that are invariable are parameters that vary across languages. UG is “taken to be a characterization of the child's prelinguistic state” (Chomsky, 1981). UG is postulated as an innate language facility that limits the extent to which languages can vary. It is the system of principles, conditions, and rules that are elements or properties of all human languages. Most UG-based researches are conducted within the framework of principle and parameter theory.

A considerable amount of the current literature on second language acquisition is concerned with the problem of whether UG is accessible or whether parameter resetting is possible. There is a continuum of views on the topic, which could be represented schematically as follows (Fernandez, 1998).Figure 2.1　Access to UG in L2 Acquisition（Fernandez, 1998）

The horizontal line represents how much knowledge of UG is available to L2 learners. At the left end is theoretical position that assumes L2 learners follow a learning process similar to the process that takes place in L1 acquisition. Here UG is accessible to the fullest extent to the adult leaner of a language, and parameters will be reset upon exposure to enough input. At the opposite end of the continuum (on the far right) stands the theoretical alternative to the UG view: UG is not accessible at all, and parameter resetting is impossible.

The positions between these two extremes are intermediate positions: learners have limited accessibility to UG, and therefore learners cannot achieve full success in L2 as their L1. UG may become inaccessible later in life according to the Critical Period hypothesis, or UG may only be accessible via the L1 parameter settings, or UG may not be accessible to adult L2 learners as it is to child L1 learner because the processing strategies employed by adult learners of a given language may not be those best suited for developing the underlying grammar of the target language.

Most second language researchers who adopt the Universal Grammar perspective assume that the principles and parameters of UG are still accessible or may be accessible to the adult language learner. However, the results of the empirical studies are mixed and indeterminate. For example, studies of L2 learners' access to the Subjacency Principle and the pro-drop parameter do not provide clear evidence to support the hypothesis. Generally, the results to test the predictions fall some way between the alternatives which the three assertions above would lead one to expect. While some predictions have been borne out, others have not, with discrepant results usually generating the need to provide post-hoc explanations (Cook, 1994; White, 1989; Gass & Schachter, 1989). As a result, judgments have to be made to “rescue” the starting hypotheses (Skehan, 1999).

More recently, as Chomsky reformulated Government and Binding theory into Minimalist Program, it is inevitable that the direction of SLA research is influenced. Scholars are no longer entangled in the issue of whether UG is accessible to L2 learners. In its place, three areas are becoming the emphases of research:

1) Within the minimalist framework, the lexicon assumes great importance. Parameterization is no longer in the syntax, but in the lexicon. Most of the constraints on language described earlier in terms of complex principles and parameters now fall out of a handful of general constraints on movement and the specific information stored in the lexicon of individual languages. Parameters are part of the lexicon and language learning is largely lexical learning.

2) The focus of research had also shifted to an investigation of functional categories which had by now been postulated as the locus of parametric differences. Functional categories can be thought of as grammatical words that in a sense form the glue of a sentence. Examples of functional categories are determiners (a, the, our, my, this), complementizers (if, whether, that), and grammatical markers (past tense endings, case markings, plural endings, and gender marking).

3) More subtle and complex explanations were being formulated to account for development in L2 acquisition, i.e., the language acquisition process. Since UG can only account for the representation of language-particular core grammars, it cannot account for the process of how the grammar is constructed. UG knowledge must interact with a learning mechanism to allow convergence on a particular representation of grammar. The learning mechanism encompasses a language parser to process the input, and learnability theory, i.e. a theory of how the input interacts with UG knowledge to allow convergence (Klein & Martohardjono, 1998).

UG-based research is strong in describing the complex structural characteristics of interlanguages, but weak in explaining the process of L2 acquisition. It focuses on the importance of input, viewing input as the trigger of language acquisition. Here output is the result of acquisition, playing little role in language acquisition. It is good at describing a formal, underlying competence, but it is less convincing with second language learning, with real-time communication, and with the relationship between performance and change. It is a property theory, and meets the need of theoretical framework condition (i.e. An SLA theory is explanatory only if it can account for each state of a learner's competence within the framework of a satisfactory property theory of linguistic competence, as defined by that theory). But an SLA theory must meet both of the theoretical framework condition and mechanism condition in order to be accepted as explanatory (Gregg, 2001). The other criterion—mechanism condition states that an SLA theory is explanatory only if it includes a transition theory that has a mechanism or mechanisms to effect changes of state in L2 competence. In other words, we require a transition theory in dealing with input and the learning mechanism that operates on that input to produce linguistic competence. For a more adequate explanation of second language acquisition, we need to resort to psychological theories.

2.3 Piaget's view of language acquisition

Although Piaget was a biologist and a psychologist, his ideas have been influential in the field of first and second language acquisition studies. In fact he studied the overall behavioral development in the human infant. But his theory of development in children has striking implications for language acquisition.

Piaget was one of those psychologists who view language acquisition as a case of general human learning. He has not suggested, however, that the development is not innate, but only that there is no specific language module. Piaget's view was then that the development (i.e., language acquisition) results mainly from children's interaction with their environment. What children learn about language is determined by what they already know about the world. Children's L1 acquisition is accompanied by the development of cognition. Piaget outlined the course of intellectual development as follows:

- The sensorimotor stage from ages 0 to 2 (understanding the environment)

- The preoperational stage from ages 2 to 7 (understanding the symbols)

- The concrete operational stage from ages 7 to 11 (mental tasks and language use)

- The formal operational stage from the age 11 onwards (dealing with abstraction)

Piaget observes, for instance, that the pre-linguistic stage (birth to one year) is a determining period in the development of sensory-motor intelligence, when children are forming a sense of their physical identity in relation to the environment.

The claim that language development is an outgrowth of other cognitive skills is labeled constructivism. According to constructivist theory, acquisition is seen as a product of the complex interaction of the linguistic environment and the learner's internal mechanisms, with neither viewed as primary. Proponents of constructivism hold the view that input does have a determining function in language acquisition, but only within the constraints imposed by the learner's internal mechanisms.

From Piaget's perspective, cognition must be considered an adaptive function. Adaptation refers to a state of organisms or species that is characterized by their ability to survive in a given environment. Cognitive change and learning take place when a scheme (cognitive structure), instead of producing the expected result, leads to perturbation (disappointment), and perturbation, in turn, leads to accommodation that establishes a new equilibrium. Learning and the knowledge it creates, thus, are explicitly instrumental. Assimilation is the process whereby changing elements in the environment become incorporated into the structure of the organism. At the same time, the organism must accommodate its functioning to the nature of what is being assimilated. When a child acquires the meaning of a word, the child's meaning of that word is made up exclusively of elements which the child abstracts from his own experience. During this process, the child must individually construct the meaning of words, phrases, sentences, and texts. Learning is the product of self-organization. As Piaget put it, “Intelligence organizes the world by organizing itself”. The most frequent source of perturbations for the developing cognitive subject is the interaction with others. This is the reason why constructivist teachers have been promoting “group learning”, a practice that lets two or three students discuss approaches to a give problem.

Proponents of constructivism look for properties of human language that have what they consider to be plausible similarities to properties of other mental operations and try to support the constructivist position by looking how these related linguistic and non-linguistic properties develop over time. They propose the concept of precursors, which means that some mental abilities are the basis for the development of other abilities. The form of the argument is generally along the lines: at stage X, a child has cognitive skill A; at stage X+1, the child acquires linguistic skill B; therefore, cognitive skill A is a precursor to linguistic skill B. The latter arises out of the former. For example, conservation may be a prerequisite for the development of linguistic structures. Children who pass tests of conservation in general use more complex linguistic structures, including the comparative, than non-conserving children (Sinclair de Zwart 1967, reported in Goodluck 2000).

The relationship between language and thought poses thorny issues and questions. In fact, cognitive and linguistic developments are inextricably intertwined with dependencies in both directions. Language is dependent upon cognitive development, and language also influences intellectual development (see the latter section about Vygotsky's view). Cognitive skills may promote the use of structures and forms already present in the child's linguistic system, and that cognitive development may enhance linguistic development in the sense of providing the child with the kind of concepts that increase the use of certain constructions. On the other hand, words shape concepts, and dialogues between parent and child or teacher and child serve to orient and mediate children's mind.

In Piaget's view of language acquisition, the cognitive development and language acquisition result mainly from children's interaction with their environment. Thus, the role of output and dialogue plays an important role in language acquisition. Since his interest lies mainly in L1 acquisition and the relationship between language and thought, Piaget's theory is not tested or applied in SLA field.

2.4 Skehan's cognitive approach

In his monograph “A Cognitive Approach to Learning Language”, Skehan (1999) proposed an information-processing account of second language learning. The fundamental concepts and claims in this book are as follows:

1）Second language learning is cognitive

In cognitive approach, it is assumed that a critical period for language development exists. For adult learners who have passed the critical period, language development can be viewed as an example of the human-information processing system at work, in a way which resembles learning in other domain, such as a computer language, or a musical instrument, or some new area of knowledge. In order to explain SLA for adult second language learners, we must search for cognitive abilities which support the different aspects of second language learning. These abilities are best understood if they are linked to the stages in which information is processed.

Drawing on the research findings made by Schmidt (1990), Long (1989), Swain (1995), and VanPatteen (1996), Skehan build an information-processing model which explain psycholinguistic processes in language use and language learning (see Fig.2.2). In this model, there are different stages in information-processing: input, central processing, and output. And within central processing, there are three separate stages of memory: sensory register, working memory, and long-term memory. The information-processing models are helpful in separating the different stages concerned, and in providing an organizing framework for more detailed discussion of the functioning of each separate stage. In this model, Skehan argues for the central importance of noticing as a trigger for interlanguage change, and the coexistence of rule- and exemplar-based representational systems.

2）Dual-coding and memory are criticalFigure 2.2　A cognitive model of language learning (Skehan, 1999)

The representation in the brain is a dual mode of processing, in which there is evidence for both rule-based systems and exemplar-based systems. The rule-based system is likely to be parsimoniously and elegantly organized, with rules being compactly structured. Such a rule-based system is likely to be generative, with rules being creative in their application, and so precise in the meanings that they can express. It is also likely to be restructurable, with new rules replacing or subsuming old rules, and then functioning efficiently as an extended system. But of course, all these gains are achieved at one considerable cost: their operation will lead to a heavy processing burden during ongoing language use. The exemplar-based system operates in more or less the opposite manner. It is heavily based on the operation of a redundant memory system in which there are multiple representations of the same lexical elements. The former emphasizes representation at the expense of processing, while the latter does the reverse. The former leads to the development of an open, form-oriented system, while the latter emphasizes meaning, and is less appropriate for underlying system change.

The two systems work harmoniously together. Normally (native) language users are adept at combining rule-based and exemplar-based modes, shifting between them as communicational circumstances make appropriate. If performance demands are high, it is likely that exemplar-based system will become a more important mode, while if performance conditions are not so pressing, and/or if precision in communication is important, the rule-based system mode will dominate.

What Skehan wanted to emphasize is that memory is critical in language use and language learning. Much of the time we rely, during rapid communication, on large chunks of memorized language, because they ease processing demands and enable real-time communication to proceed. Memorized language such as lexical phrases, formulaic expressions, lexicalized sentence stems, phrase schemata and sentence schemata are much more important than was previously thought.

3）Tensions between learning and performance

According to Skehan, post-critical period learners emphasize meaning, so there is a danger that second language learners will rely too heavily on the exemplar, memory-based system. This is potentially harmful for second language development because complex form may be avoided. As such, there exist tensions between form and meaning (fluency), first of all, and then within form between accuracy and complexity-restructuring. Given this potential state of affairs, the challenge facing pedagogy is to establish principles which enable instruction to foster balanced development.

The information-processing framework provides us with a developmental view of the acquisition of skills in terms of both knowledge (mental representation) and executive control (the processing of mental representations). The framework seems especially designed to account for the process of language use and language acquisition, and regards language use and language acquisition as constrained by the operations of a limited capacity information-processing system. The notion of dual-code representation is enlightening here, but dual processing has become a hot issue recently. Perhaps Skehan should be defending himself against connectionist one-system models which claim that rules are unnecessary. What's more, despite the importance of memorized language, the framework does not deal adequately with vocabulary acquisition, a vast area of current research.

Although Skehan recognized and included output as a component of information-processing model, he did not consider it as an important component of language acquisition. This is the problem with his model, which we will discuss in next chapter.

2.5 Gass's integrated model

In Skehan's cognitive approach, language use and language learning are considered to follow the information-processing model which has separate stages of memory (sensory register, working memory, and long term memory), and these stages are helpful in making sense of the complex phenomenon of memory. An alternative to the stage model is levels of processing model (see Lahey, 2001). The latter model suggests that the distinction between working memory and long-term memory is a matter of degree rather than separate stages. Information will be kept only briefly if it is processed at a shallow level, but it will be kept much longer if it is processed at a deeper level. Thus the differences between working memory and long-term memory are not, in this view, differences between two different memory systems operating according to different principles. Rather, these differences are the results of different levels of processing during the encoding process. Furthermore, there is a continuum of levels of processing, ranging from very shallow to very deep, rather than just two types of storage (short and long).

In line with levels of processing model, Gass (1988, 1997, 2001) proposed a model of second language acquisition (Figure 2.3). Five stages are proposed to account for the conversion of input to output: apperceived input, comprehended input, intake, integration, and output. These stages are related to the depth of processing of the input. The factors which mediate between one level and another are also dealt with in the model.

The model described here represents a dynamic view of the process of acquisition. The first stage of input utilization is the recognition that there is something to be learned, that is, that there is a gap between what the learner already knows and what there is to know. This is called apperception. The next level in the process of acquisition is comprehended input. Comprehension in this sense represents a continuum of possibilities ranging from semantics to detailed structural analyses.Figure 2.3　A model of second language acquisition (Gass & Lelinker, 2001)

Comprehended input is in a deeper level of analysis than apperception. The third stage, intake, is the process of assimilating linguistic material; it refers to the mental activity that mediates input and grammars. It is where generalizations are likely to occur, and where memory traces are formed. After the intake component has performed its task of processing the input and matching it against existing knowledge, there are at least two possible outcomes, both of which are a form of integration. One is the development per se of one's second language grammar, and the other is storage. The fifth stage is output. Output plays an active role in acquisition in that it serves as a means of hypothesis testing and forces a syntactic analysis of language. It is also an overt manifestation of the acquisition process. Like Skehan, Gass also put much emphasis on attention, awareness and consciousness. In addition, Gass's model shows the multiple roles that language transfer and universals have. Other mediating factors like personality and affect are also dealt with in the dynamic and interactive way.

The model is intended to reflect the dynamic and interactive nature of acquisition. And deep levels of processing leads to change and development of interlanguage. Here memory is a constructive process rather than a storehouse. Remembering appears to be far more decisively an affair of construction rather than one of mere reproduction. Thus, knowledge requires an active approach, not a passive one. One does not simply reproduce it; one constructs it. Interaction is important in that it is not only a medium of practice; it is also the means by which learning takes place. In other words, conversational interaction in a second language forms the basis for the development of syntax; it is not merely a forum for practice of grammatical structures.

One problem with this model is that although it include both cognition and UG in the model, it is not clear how they interact with each other. Because either the cognitive or UG-based views of SLA cannot accept the other's point of view, we wonder how Gass's model moves towards a compromise, with revision on both sides to take into account findings in both disciplines.

Gass provided us a comprehensive model of second language acquisition within cognitive framework. In this model, output plays an active role in acquisition in that it serves as a means of hypothesis testing and forces a syntactic analysis of language. It is also an overt manifestation of the acquisition process.

2.6 Connectionism and emergentism

Within cognitive psychology, there are two different accounts of how knowledge is organized and learned in our brain. One is symbolism; the other is connectionism (or Parallel Distributed Processing, PDP). Previous models are symbolic in nature. Symbolic processing involves the manipulations of representations of data, often referred to as “patterns” or “rules”. Connectionist models however, insist that there are no concrete symbols and rules as such; the entities that a connectionist system uses to characterize the world are fluid patterns of activation across portions of a network. Symbolic models often feel it is enough to describe a point in this process, whereas connectionist models try to see process from one state to another. In the connectionist model, information processing involves the activation of the requisite interconnections; learning arises when, as a result of experience, the strength of the connections (i.e. “weights”) between units is modified. All knowledge is embodied in a network of simple processing units joined by connections which are strengthened or weakened in response to regularities in input patterns.

The connectionist model resembles other cognitive models in that the model also consists of sensory register, working memory and a storage area called the network/ store. The model is illustrated in Figure 2.4 (Waring, 2004).Figure 2.4　A representation of the interface between input data, working memory and memory networks (Waring, 2004).

In this model, input is received at the sensory level (the level at which information is registered on the retina or ear drums). The central executive will attend to some of this input and put it into working memory. The information considered most salient and unexpected is usually that which is attended to. This input then becomes conscious in that you are consciously (at varying levels) aware of it -- the rest of the input is not attended to and is effectively discarded. This input is then compared with pre-existing lexical information in the network which, in keeping with schema theory, is in a constant state of expectation that the incoming data will match pre-existing vocabulary knowledge and thus confirm that knowledge and lead to comprehension.

Working memory plays a vital role in this model. Working memory has three major components: the central executive, the phonological loop, and the visuo-spatial sketchpad (Baddeley, 1990). The central executive regulates, monitors and coordinates the operation of the other components. The phonological loop is divided into the articulatory control system, which can work at the sub-vocal level, and a phonological store which holds speech based information. The final part is the visual-spatial sketchpad which receives inputs either directly from visual perception or by retrieving information from long-term memory in the form of images. The central executive functions as a mediator between the network and the input. It is primarily concerned with (among other things) what to attend to, what to ignore, how to put this into the network and so on. There is freedom of movement between the network and working memory and the higher cognitive functions allowing the flow of information back from the network into working memory for re-evaluation and reflection.

Emergentist views of language acquisition hold that it is primarily these systems that the child uses in bootstrapping their way into language (N. C. Ellis, 2001). Connectionism provides a set of computational tools for exploring the conditions under which emergent properties arise. Like other theories, including connectionist approaches, functional linguistics, constructivist approach, and cognitive linguistics, emergentism denies any innate linguistic universals. “There is no language acquisition device specifiable in terms of linguistic universals, principles and parameters, or language-specific learning mechanisms. Rather, language is cut of the same cloth as other cognitive processes, but it is special in terms of its cognitive content.” (N. C. Ellis, 2001) It is believed that language is learned. Language learning results from general processes of human inductive reasoning being applied to the specific problem of language. Emergentists believe that simple learning mechanisms, operating in and across the human systems for perception, motoraction and cognition as they are exposed to language data as part of a communicatively-rich human social environment by an organism eager to exploit the functionality of language, suffice to derive the emergence of complex language representations. Emergentist approach to language acquisition believes that as the study of language turns to consider ontogenetic acquisition processes, it favors a conclusion whereby the complexity of the final result stems from simple learning processes applied, over extended periods of practice in the learner's lifespan, to the rich and complex problem space of language evidence.

There are now many separate connectionist simulations of the acquisition of morphonlogy, phonological rules, novel word repletion, prosody, semantic structure, syntactic structure, etc. (e.g., Levy, et al., 1995; MacWhinney & Leinbach, 1991; Rumelhart & McClelland, 1986). These simple demonstrations repeatedly show that connectionist models can extract the regularities in each of these domains of language, then operate in a rule-like (but not rule-governed) way. Linguistically-relevant representations are an emergent property of the network's functioning.

Proponents of emergentism focus on the language acquisition process rather than language acquisition device. In this process, chunking is the overarching principle. Chunking is a major technique for getting and keeping information in short-term memory. “A chunk is a unit of memory organization, formed by bringing together a set of already formed chunks in memory and welding them together into a larger unit. Chunking implies the ability to build up such structures recursively, thus leading to a hierarchical organization of memory. Chunking appears to be a ubiquitous feature of human memory. Conceivably, it could form the basis for an equally ubiquitous law of practice.” (Newell, 1990) The power law of practice describes the rate of acquisition of most skills.

Another theory with similar orientation is MacWhinney's Competition Model. MacWhinnery (1989) finds that connectionism offers a powerful formal framework that correctly expresses the processing and learning claims of the Competition Model. To characterize the cognitive abilities of the learner, the Competition Model relies on findings from connectionism. The connectionism emphasizes the roles of transfer, automatization and parasitism in learning of the L2. So does the Competition Model. This is not surprising, perhaps, given that both connectionism and the Competition Model share a number of features in common, such as the notions of “network” and the “weight” of cues or connections.

The Competition Model (MacWhinney, 1987, 1997) sees all of language acquisition as the process of acquiring from language input the particular cues which relate phonological forms and conceptual meanings or communicative intentions, and the determination of the frequencies, reliabilities and validities of these cues. This information then serves as the knowledge base for sentence production and comprehension.

Very little work on connectionism has been done in second languages. This lack of work does not mean a lack of interest however and is understandable in that the field is only 10 years old. Most of the research has been done in the first language and it has only been very recently that work has started on a second language.

Connectionists and emergentists view language as a complex, dynamic system and language use/acquisition as dynamic adaptedness to a specific context. This proves a useful way of understanding L2 system development. Language output, hence, is the process in which the language resources of each individual are uniquely transformed through use.

2.7 Vygotsky's zone of proximal development

Vygotsky was a psychologist but his studies on conscious human behavior led him to investigate the role that language plays in human behavior. Vygotsky's point of view is simply that social interaction plays an important role in the learning process. He places an emphasis on the role of “mediation” and “zone of proximal development” in the development of thought and language. His theory of psychology is called social constructivism.

The most fundamental concept of Vygotsky's theory is that the human mind is mediated. Vygotsky (1978) conceived of the human mind as a functional system in which the properties of the natural brain are organized into a higher mind through the integration of symbolic artifacts into thinking (examples of symbolic artifacts are numbers and arithmetic systems, music, art, and above all language). Higher mental capacities include voluntary attention, intentional memory, planning, logical thought and problem solving, learning, and evaluation of the effectiveness of the processes.

The site where social forms of mediation develop is the “zone of proximal development”. Zone of proximal development is the difference between the child's capacity to solve problems on his own, and his capacity to solve them with assistance. In other words, the actual developmental level refers to all the functions and activities that a child can perform on his own, independently without the help of anyone else. On the other hand, the zone of proximal development includes all the functions and activities that a child or a learner can perform only with the assistance of someone else. The person in this scaffolding process, providing non-intrusive intervention, could be an adult (parent, teacher, caretaker, language instructor) or another peer who has already mastered that particular function.

According to Vygotsky, all fundamental cognitive activities take shape in a matrix of social history and form the products of sociohistorical development. That is, cognitive skills and patterns of thinking are not primarily determined by innate factors, but are the products of the activities practiced in the social institutions of the culture in which the individual grows up. Consequently, the history of the society in which a child is reared and the child's personal history are crucial determinants of the way in which that individual will think. In this process of cognitive development, language is a crucial tool for determining how the child will learn how to think because advanced modes of thought are transmitted to the child by means of words.

To Vygotsky, a clear understanding of the interrelations between thought and language is necessary for the understanding of intellectual development. Language is not merely an expression of the knowledge the child has acquired. There is a fundamental correspondence between thought and speech in terms of one providing resource to the other; language becomes essential in forming thought and determining personality features.

According to Vygotsky, an essential feature of learning is that it awakens a variety of internal developmental processes that are able to operate only when the child is in the action of interacting with people in his environment and in cooperation with his peers. That is why his theory is classified into the interactionism.

In conclusion, Vygotsky contends that language is the key to all development and words play a central part not only in the development of thought but also in the growth of cognition as a whole. Within this framework, child language development, thus acquisition, can be viewed as the result of social interaction. Compared with Piaget's view of language acquisition (see above), Vygotsky's theory is more social in orientation. Output, activity and interaction are important key words in this theory. Sociocultural theory of SLA can be viewed as its development in second language acquisition research.

2.8 Sociocultural view of language acquisition

Current conceptualizations of sociocultural theory draw heavily on the work of Vygotsky (1986), as well as later theoreticians. Vygotsky viewed language acquisition as the result of social interaction. Although second language research within Vygotskian theory has been carried out for decades in the former Soviet Union, such research has only minimal influence on L2 research in the west. There are two principal reasons. First, only recently have scholars working outside of the former Eastern bloc begun to have fuller access to the writings of Vygotsky and his followers. Second, Vygotsky turned to philosophy to explain higher order mental functions like thinking and consciousness. Vygotsky proposed that since thinking is the function of the cerebral organ, the explanation of the process is not to be found in the internal structure of the organ, but in the interaction between thinking bodies (humans) and between thinking bodies and objects (humans and socioculturally constructed artifacts). This is in stark contrast with second language acquisition research in the west, which has squarely situated itself within the natural science research tradition—a tradition that values predictive explanation and controlled, heavily quantitative experimentation.

Indeed, higher order mental abilities are of great interest to educators and researchers in SLA circles. When investigating SLA process and learners, we are not only concerned with the development of linguistic abilities, but also the development of learners' cognition and affective factors. In some sense, learners' cognition and affective factors are becoming more and more important. In our increasingly complex and specialized society, it is becoming even more imperative that individuals are capable of thinking divergently and creatively. It is also important that individuals see the relationships between seemingly diverse concepts. The ability to construct lessons which include higher order questioning is part of the new curriculum. “The Outline of Curriculum Reform of Elementary Education” (2001) clearly stated that the goal of the reform is to develop a whole person in every student, to foster creativity and practical abilities. Our hope is that second language researchers will begin to explore the potential that sociocultural theory and Vygotskian research methodology have for developing an even fuller understanding of second language phenomena.

The SLA research within sociocultural theory mainly deals with the fundamental concepts of mediation, inner speech and activity theory (Lantolf, 2000). According to sociocultural theory, dialogue in a learning setting plays an important part in helping learners to internalize or mediate ideas and knowledge from the social plane. Learning advances when tasks are pitched just beyond the learners' zones of proximal development, i.e. the area between students' actual or unassisted performance and their performance when provided with outside support or assistance. In order to advance the learner towards more complex forms of understanding, scaffolding can be provided by peers and others. Whereas much of the research applying Vygotsky's work has been based on the asymmetric interactions of teachers and learners, contemporary research is also investigating the interactions in more symmetrical learning environments involving learners working collaboratively. Thus, the interactions that occur among peers are legitimate forms of scaffolding that offer opportunities and support for cognitive development. When learners have to explain ideas to each other, irrespective of the relative abilities of those involved, a more explicit and organized understanding can result. This form of co-construction leading to cognitive change is critical to the development of higher order thinking processes.

The function of inner and private speech in second language learning and performance is anther focus in this framework. Large-scale survey of ESL (English as a second language) learners was designed to elicit information on their use of inner speech in the second language acquisition process. And analysis of inner speech may give us new insights into the nature of the second language learning process. Frawley and Lantolf (1985) claim that as learner' proficiency increases their use of private speech decreases. It is also found that cultural background of L2 speakers may override proficiency level with respect to the frequency of private speech production. Ushakova (1994) reports on the findings of an experimental study of lexical acquisition in a second language, from which she concludes that the inner speech patterns laid down during first language acquisition serve as hosts onto which second languages are mapped.

Another important theme within this framework is Activity Theory. According to Vygotsky, activity is a mental process, a frame, a sociocultural interpretation, in which the events are constructed by the participants based on their context of appearance. When a participant is engaged in an activity, she/he “is functioning in some socioculturally defined context”. The Activity Theory comprises three levels: activity (linked to motive), action (linked to goal), and operations (which determine the means). The main idea of the Activity Theory is that, human sociocultural activity that gives rise to higher forms of cognition, is comprised of contextual, and circumstantial dimensions. The motive and goal constitute a “kind of vector”, determining the direction and amount of effort an individual exerts in carrying out the activity. Motives “energize” an activity and goals impart directionality. The actual realization of the activity is achieved through specific material circumstances at the operational level. Thus the level of motive answers why something is done, the level of goal answers what is done, and the level of operations answers how it is done. The link between socioculturally defined motives and concrete actions and operations is provided by semiotic systems, of which language is the most powerful and pervasive ( see Lantolf & Appel, 1994).

Research from the perspective of activity theory mainly focuses on how language acquisition and activities are intertwined with each other. Ahmed (see Lantolf & Appel, 1994), for example, analyzes and compares how L2 and L1 speakers employ language in problem-solving activity. He focuses on bringing to light how apparently aberrant linguistic patterns play an important role in a speaker's attempt to mediate mental activity in order to complete a difficult puzzle task. Coughlan and Duff (see Lantolf & Appel, 1994) show that tasks cannot be designed to elicit specific samples of interlanguage data independently of the speaker who engages in communicative linguistic activity. Their L2 protocols support Vygotsky's argument that speakers are agents active in controlling their environment; consequently, tasks cannot be predetermined, but emerge from the interaction of speakers, settings, motives, and histories.

Recent research on task-based approach, cooperative learning, and self-regulated learning is linked to Vygotsky's Activity Theory. Tasks are such activities which require learners to use language, with emphasis on meaning, to attain an objective. Cooperative learning focuses on the social interaction between peer learners and between the learner and the teacher. Our ultimate goal in language teaching is to cultivate a self-regulated learner who takes the responsibility for his/her own learning and control his /her own behavior.

The sociocultural perspective has profound implications for teaching, schooling, and education. A key feature of this emergent view of human development is that higher order functions develop out of social interaction. Based on this social interactionist view, Williams and Burden (1997) constructed a social constructivist model of language teaching and learning. The model emphasizes the dynamic nature of the interplay between teachers, learners and tasks, and provides a view of learning as arising from interactions with others.

The socialcultural view of language acquisition is enlightening in that it is possible to cultivate higher order thinking in language teaching and learning. Traditional teaching often employs a lock-step approach to the acquisition of skills, with lower order skills being taught first. There is a belief that lower order skills are prerequisites to higher order skills and that mastery of lower order skills automatically leads to higher order skills. In fact, lower order and higher order skills can be taught concomitantly, with students mastering both levels as they apply their learning, rather than learning the skills, practicing and then applying them. This prevents skills from being learned in isolation and students having to relearn how to apply them to real world tasks. The students are active in the learning process, applying problem-solving in context which, in turn, aids in the acquisition of skills, giving the students a reason to learn and helping them to learn.

The socialcultural view of language acquisition put much emphasis on language output and interaction. Yet its research method is mainly qualitative such as conversation analysis, thus their results and findings are not fully accepted by the mainstream researchers. And postgraduate students still hesitate to adopt such a paradigm in their thesis because of the research method.

2.9 Summary

In this chapter, we examine theories of second language acquisition research and the role of output in these theories. Each of these models accounts persuasively for what it considers the crucial aspects of L2 learning. Each of them is at best a piece of the jigsaw (Cook, 2000). Do the pieces add up to a single picture? The answer is probably yes. At the moment there is no overall framework for all the models. When they are fitted together, an overall model of L2 learning will one day emerge. At the moment there are many area-specific models, each of them providing some useful insights into its own province of L2 learning.

The review of these theories suggests that earlier claims, by Krashen and UG-based SLA researchers, for example, emphasizes the crucial role of input and the innate factors in acquiring second language. In these theories, output is the result of acquisition, not the cause of acquisition, and little emphasis is given to output. Later theories within cognitive framework like Skehan's theory and Gass's integrated theory recognize the role of output in language acquisition, but fail to attatch enough importance to language output. Other factors, such as noticing, depth of processing and memory as well as input are the focus in these theories. Only recent theories such as connectionism and sociocultrual theory hold that relationship between second language use (mainly output) and second language development is more direct, and the role of output is crucial in second language acquisition. The meaning of “output” has developed, too. The word “output” was used to indicate the outcome, or product, of the language acquisition device. Output was synonymous with “what the learner has learned” in the 1980s. There has been a shift in meaning from the 1980's to now from output being understood as a noun, a thing, a product to output being understood as a verb, an action, a process (Swain, 2008).

With regard to the influence each of these theories has on SLA, the difference can be found in the general emphasis: linguistics focuses on the products of acquisition (i.e. a description of the system produced by learners), cognitive perspective focuses on the process by which those systems are created (e.g., a description of the process of the way in which learners create learner systems), and social constructivism focuses on the social factors and cultural factors in language acquisition and development of mind. One can find that linguistics-based research has limited power, because it focus on the result and representation of language acquisition, and neglects the process of acquisition. The cognitive perspective is the prominent paradigm in the field, with the traditional positivist paradigm. However, sociocultural perspectives have become an alternative paradigm.

Chapter 3 Language Output and Information-Processing Framework of SLA

This chapter sketches out a conceptual framework depicting the roles of output in language acquisition and proposes a new information-processing framework of SLA. It first points out the problems with Skehan's cognitive approach in the literature, and examines the underlying psycholinguistic rationale of comprehension and production. Following recent theories such as connectionism and sociocultrual theory, the chapter argues for the central importance of output in promoting language performance and cognitive development of the learner within information-processing framework.

Here Skehan's cognitive approach or information-processing framework is chosen as the theoretical basis. There are several reasons for choosing this theoretical basis. For one thing, experimental studies within the information-processing framework seem particularly fruitful, and it is the mainstream perspective in SLA field. For another thing, cognitive approach has a great many implications to and applications in language learning and teaching, including task-based language teaching and learning which is feasible in second language classroom.

3.1 Skehan's cognitive approach and its problems

Skehan's information-processing framework (see Chapter Two, 2.4 ) provides us with a developmental view of the acquisition of skills in terms of both knowledge (mental representation) and executive control (the processing of mental representations). The framework seems especially designed to account for the process of language use and language acquisition and regards language use and language acquisition as constrained by the operations of a limited capacity information-processing system. This model explicitly addresses issues of representation, learning (change), and processing performance (Liu Chunyan, 2004). And it also provides us with operationizable framework to build hypothesis and test hypothesis of empirical studies. It provides us an account of the process of second language acquisition and teaching process. When we gain more insights in language acquisition and learning, it is clear that language use is as important as language knowledge, and the process of language learning is more important than language learning result. Not surprisingly, it is most prevalent theory in SLA.

However, there has been a lot of debate among scholars about some of the notions in this model. Some scholars challenge the notion of limited capacity of attention. They hold a multiple-resources view of processing — that is, that learners, like native speakers, have the capacity to attend to more than one aspect of language and language processing at the same time. Robinson (2001, 2003), for example, advocates two propositions: 1) that attentional resources are not limited in the way Skehan and Foster (2001) argue, but instead learners can access multiple and non-competing attentional pools, and 2) that, following Givon (1985), complexity and accuracy in a task correlate, since they are each driven by the nature of functional linguistic demands of the task itself.

Another point of debate is the separate stages of memory (sensory register, working memory, and long term memory) in Skehan's model. An alternative to the stage model is levels of processing model (see Lahey, 2001). The latter model suggests that the distinction between working memory and long-term memory is a matter of degree rather than separate stages. Information will be kept only briefly if it's processed at a shallow level, but it will be kept much longer if it is processed at a deeper level. Thus the differences between working memory and long-term memory are not, in this view, differences between two different memory systems operating according to different principles. Rather, these differences are the results of different levels of processing during the encoding process. Furthermore, there is a continuum of levels of processing, ranging from very shallow to very deep, rather than just two types of storage (short and long). In line with levels of processing model, Gass (1988, 1997, 2001) proposed a model of second language acquisition. Five stages are proposed to account for the conversion of input to output: apperceived input, comprehended input, intake, integration, and output. These stages are related to the depth of processing of the input. Gass' model goes further in acknowledging the stage of output as the final stage in language acquisition. However, this model has its limitations (see Chapter 2).

The most contradictory point in Skehan's theory is that although it acknowledges the importance of language use/output and language learning process instead of language knowledge and learning result, language use/output does not have a place in the language acquisition system. According to Skehan, noticing the input and dual-coding in the memory play a central role in language acquisition and language performance. Skehan did mention Swain's output hypothesis, and summarized 6 roles for output (1999:16-19). However, he thought that it is doubtful whether actual output favors form or emphasizes fluency at the expense of form. He argued that in real-time communication, reliance on communication strategies seemed to be harmful to linguistic development. He concluded that language use, in itself, does not lead to the development of an analytic knowledge system since meaning distracts attention from form (p27). Clearly, he recognizes the dual code system of rule-based and exemplar-based systems, but neglects how they interact each other. He emphasizes the tension between learning and performance, but neglects the fact that performance is a crucial part of learning.

When discussing the relationship between the dual-mode system, Skehan recognizes that “neither the rule-based nor the exemplar system is ideal separately…the two systems might work harmoniously together.” (ibid, p89). Drawing on the first language acquisition research, Skehan assumes that L2 learners experience the developmental sequence of lexicalization — syntacticalization — relexicalization. The initial stages are primarily lexical in nature. Contextually coded exemplars are used to communicate meanings in a direct manner. Then, at a later stage, strong process of syntacticization comes into play and operates upon the lexical repertoire that the child has developed. The result is that language which was available on a lexical base now becomes reorganized to be syntactically based, and the benefits of such a rule-governed system come into play. What happens in subsequent development is that language which has been syntacticized is then relexicalized for the access as units for processing with minimal computational demands. The sequence can be represented as in Figure 3.1.Figure 3.1　Sequence of exemplars of lexical items (from Skehan, 1999)

It is assumed, in the first language case, that the syntacticalization process is “due to the maturation and operation of an inbuilt propensity to process language input in a manner qualitatively different from the way other material is processed, and that it is the operation of a LAD which produces this state of affairs (ibid, p90)”. In the case of L2 acquisition, when the critical period is past, this automatic engagement of a system designed to syntacticize is no longer an option. “As a result, the internally-generated pressure for syntactization will not come into play. In other words, there is a danger that the second language learner will not progress beyond the first of the three stages mentioned above.”

Drawing on Klein (1986), Skehan then proposes a system for second language learners, i.e. cognitive processes such as matching, analysis and synthesis are necessary to contrive the movement through all three stages. “Matching” means noticing the gap between one's current language system and the language one encounters (Klein, 1986). It is important for the learners to analyze the linguistic units they are using, so that they can access this same material as a rule-based system. Equally, it is important that when material does become available as such a system, learners should engage in the complementary process of synthesizing such language so that it will then become available in exemplar, memory-based form as well. But Skehan does not address the issues of how and when these processes will take place.

In my opinion, it is output and production which will engage learners in such processes. During language production, learners will “notice the gap”, realizing where their interlanguage system is inadequate. And during production, learners are pushed to examine syntax and even analyze the language to be used metalinguistically. Also, during the production, the learner has to synthesize the language in such a way as to meet the communication pressure. All these coincide with the functions of output put forward by Swain. Through empirical studies, Swain and Lapkin (1997) were able to show that:

-output caused a mismatch to emerge between the language which was known and that which was needed;

-the need to express meaning pushed learners to examine syntax as a means of achieving meaning;

-restructuring, a change in the underlying interlanguage system, occurred as the mismatch between current knowledge and required knowledge was resolved.

It is Swain's output hypothesis that we will examine in more detail in next section.

3.2 Output: result or cause of acquisition

Swain's output hypothesis (see Chapter 1) has generated many empirical researches into the roles of output in SLA. These studies have reported positive and promising, though not unconditional, findings for the specific functions of output. Izumi and Bigelow (2000) examined the noticing function of output through essay-writing tasks and text reconstruction tasks. The results indicate no unique effects of output. But they find that extended opportunities to produce output and receive relevant input are crucial in improving learners' use of the grammatical structure. A closer examination of the data suggested, however, that output did not always succeed in drawing the learners' attention to the target form, a phenomenon that seems related to both learner and linguistic factors.

Swain & Lapkin (1995) used think-aloud protocols to examine whether learners “notice the gap” when problems arise in output. In this introspection study, they sought “to try to arrive at the mental processes…reflected in the changes students made to their output” (p381). The researchers examined the ability of 18 grade 8 immersion students learning French to consciously reprocess their IL output without any sort of external feedback when faced with a performance problem. The task given to the students was to write a report on some environmental problem. The students were instructed to think aloud while writing and especially when they were faced with a problem. Swain and Lapkin found that there were 190 occasions in which students encountered a linguistic problem (a gap) in their output. On each occasion, they forced themselves to modify their output toward comprehensibility or accuracy. Swain and Lapkin argued that the activity of producing the TL enabled the learners to notice a gap in their existing IL capacity. This noticing pushed them to reprocess their performance consciously in order to produce modified output.

There are also plenty of studies of fluency function (Bygate 2001; Dekeyser 1998), hypothesis-testing function (Ellis and He, 1999; Nobuyoshi and Ellis 1993; Pica 1988; Pica et al. 1989; Shehadeh 1999, 2001), and metalinguistic function (Kowal and Swain, 1994; Swain, 1995, 1998; Swain and Lapkin 2001). However, there is little literature on what processes output causes in the central processing (e.g. the two memory system and the interplay in dual-code representation system), what the role of output is in the whole cognitive system, and what the role of output is in the whole acquisition system. Swain, in her latest work, tends to resort to the sociocultural theory to explain learner interactions. Swain and Lapkin (2001) compare dictogloss and jigsaw tasks for their capacity to engage learners in collaborative interaction, and show how each learner may contribute aspects of language structure that the other cannot, and as a result, both learners in pair work would benefit. Although sociocultural theory may be an alternative explanation to output hypothesis, it would be theoretically more consistent if we can prove that output plays a role in the cognitive framework.

As for the role of output in the whole acquisition system, Tarone and Liu (1995) conducted a longitudinal study of a learner's participation in different kinds of social interactions. They concluded that learner's participation in different kinds of interactions (including output) can differentially affect the rate and route of the acquisition process. This longitudinal study is significant in that the research to date has been invariably crosssectional in design, and fails to show a direct relationship between task-design and L2 acquisition.

Another aspect of output research is the learner factors. As a living, active participant in the acquisition process (who has enhanced cognitive abilities compared with the child learning a first language), learners' affective variables such as attitude and motivation play an important role in the output process. In this aspect, Wang Chuming et al. (2000) reports on a one-semester-long experiment on improving Chinese speaking EFL learners' English by means of composition-writing. The subjects consisted of 201 English majors in their first year of study at Guangdong Foreign Studies University. They used a questionnaire to investigate whether the subjects experienced changes in their motivation, attitude and confidence in learning English after the experiment. Responses to the questionnaire showed that the subjects welcomed the new method and consequently felt more confident in their own writing ability and in their use of English.

In summary, the research on the role of output is still piecemeal and incomplete. The studies did not directly address the issue of whether output leads to acquisition or not. As Shehadeh (2002) points out, existing research was mostly descriptive in nature, focusing primarily on occurrence per se rather than acquisition or whether and how interaction/output can be a source of competence in the L2. Much of the empirical research confined itself to examining the different learner and contextual factors that affect learner's interaction (e.g., gender difference, task type, ad discourse variables) rather than investigating whether and how interaction/output can help with L2 learning. It is necessary to investigate output and production in an all-round way from the cognitive perspective. And it is also necessary to compare production process with comprehension process to see the differences between the two processes.

3.3 Comprehension process and production process

As an important process in information processing and language acquisition, speech comprehension has drawn scholar's attention since Krashen put forward the “comprehensible input hypothesis”. However, psycholinguistic research over the past decades has accumulated enough evidence to suggest that comprehension is necessary but not sufficient for acquisition to take place (see Sharwood Smith, 1986; Brown, 1995; Faerch and Kasper, 1986; Schmidt, 2001). Some general characteristics of human speech comprehension process include the following:

1) Comprehension is an active rather than a passive skill. Successful listening is a collaborative construction of a “mental model” of what is being talked about. One way of achieving such a model is by negotiating meaning to ensure understanding but another way may be through exposure to premodified input.

2) Speech comprehension is a highly complex and resourceful system. Depending on the listening purpose, the listener can rely on the schematic knowledge, contextual knowledge as well as linguistic knowledge.

3) It is now widely accepted that reading comprehension is interactive in nature, i.e. the reader uses both bottom-up and top-down processes. The interactive-compensatory model of reading (Stanovich, 1980) claims that readers use linguistic cues (referred to as a bottom-up process), and contextual clues and world knowledge (referred to as top-down process). In this comprehension system, there are compensatory mechanisms. If there is a deficiency in any particular process (e.g. weak syntactic knowledge), other processes (e.g. higher-order knowledge structures, such as contextual or general world information that the reader has access to) can compensate for the weak knowledge source.

4) Comprehension involves degrees of understanding (R. Ellis, 2003). At one end of the continuum is total non-comprehension, i.e. the listener is unable to segment the continuous stream of speech, while at the other is successful comprehension, i.e. the listener has attended to the message fully and is able to construct a coherent interpretation. Intermediate levels of comprehension arise when the listener can hear words but cannot fully understand them and when the listener is able to hear and understand but has “switched off”. A key question is what degree of comprehension is necessary for acquisition to take place. This question is not fully addressed by researchers to date.

What we are interested here is whether the processes involved in comprehending and processes involved in learning are the same or different. Here are the research findings in this aspect:

5) Comprehension is not strongly related to acquisition. Loschky's study (1994) found that interactionally modified input aided comprehension but not acquisition. In fact, the correlations between the different measures were very low and non-significant in this study. Loschky concludes that it is not possible to posit a linear relationship between comprehension and acquisition. This conclusion is in part borne out by R. Ellis (1995). He then claims that the relationship between comprehension and acquisition is complex. Comprehending a message does not guarantee acquisition of new word meanings. Also, it is not necessary to comprehend a complete message in order to acquire a new word embedded in it. Hence, R. Ellis (2003) distinguishes two different functions of listening—listening-to-comprehend and listening-to-learn.

6) For acquisition to take place, noticing plays an important role. Vanpatten (1996) assumes that attention is a prerequisite for learning to take place. He argues, however, that learners' attention tends to be drawn to certain parts of the input, particularly those that are immediately relevant to the message content. Learners may never attend to purely formal, functionally redundant forms unless some form of instructional intervention forces them to do so. Schmidt (2001) also claims that “people learn about the things they attend to and do not learn much about the things they do not attend to”.

7) The “task-induced involvement load” can be used to evaluate the effectiveness of listening-and-do tasks for vocabulary learning. Laufer and Hulstjn (2001) propose that task effectiveness is determined by the learners' level of involvement. Involvement is operationalized in terms of three constructs: a) need, a motivational factor that concerns the extent to which the task requirements create a need to process a word; b) search, a cognitive factor concerning whether learners engage in trying to find the meaning of the word by consulting a dictionary or another authority such as a teacher; and c) evaluation, a cognitive factor that relates to what learners do with a word they have encountered to establish its semantic and formal properties.

What all this implies is that the comprehension and learning/acquisition involve separate processes. Those studies based on listen-and-do tasks that examined both comprehension and language acquisition did not find a close relationship between the two. Comprehension could occur without acquisition and acquisition could sometimes take place even though learners had not fully comprehended the input. In other words, there are two ways of processing input, one involving comprehension and the other acquisition. Acquisition only occurs when learners discover that their original surface structure representation of the input does not match the semantic representation required by the situation. It will not occur if learners rely purely on top-down processing by utilizing non-linguistic input. And acquisition will benefit from comprehension process when the learner attends to the form, and when the “task-induced involvement load” is heavy.

On further thinking, one may wonder what output can do to enhance learning. In light of the predictions made by the output hypothesis as discussed earlier, it can be posited that output has the potential for noticing the mismatch between what the learners want to say and what they can say; output also has the potential for altering the manner in which learners process input (i.e., from semantic processing to syntactic processing); and output surely involves heavier “task-induced involvement load” than comprehension. How does this occur in psycholinguistic terms? What are the cognitive mechanisms involved? It is to these topics that we now turn.3.3.2　Speech production processes

This section will first introduce Levelt's Production Model. Of several psycholinguistic models of speech production proposed in the literature, Levelt's Production Model is the most influential one. This model is good at explaining how the speaker notices the gap when facing problems in their production process and how grammatical encoding and monitoring force the learner to move from the semantic processing prevalent in comprehension to the syntactic processing needed for production. Then we will try to explain how fluency is developed during production. Finally, we will proceed to describe Robinson's Cognition Hypothesis to explain how accuracy and complexity are developed in production process. Although the different theories in this section deal with different aspects of production process, they form a whole picture of production processes from the information processing perspective.

1）Levelt's speech production model

The Production model developed by Levelt (1989, 1993, 1999) assumed that lexical processing is just another form of information processing generally. The model has been adapted to account for L2 data by many L2 researchers (Bygate 2001; de Bot et al. 1997; Dornyei and Kormos 1998; Kormos 1999). A brief sketch of the model is provided in Figure 3.2.

The backbone of the human language production (and perception) system is formed by the relation among three distinct levels of representation: the conceptual level, the lemma level, and the word form level. There are five distinct components in this production model: the conceptualizer, the formulator, the articulator, the acoustic-phonetic processor and the parser. On the outgoing production basis, speech is processed roughly in the ways shown in Figure 3.2 (see the left side of Figure 3.2, moving from top down).Figure 3.2　Levelt's (1993) Production model

• The conceptual content of a speech act is planned. Concepts (ideas) are then move to the utterance formulator via message generation involving encoding into proposition.

• Propositional messages are moved to the articulator via formulation, which involves grammatical and then phonological encoding, resulting in an internal, partially encoded speech plan.

• Internally formulated utterances are moved to the environment (i.e., “produced”) via phonetic encoding and articulatory processes.

• Simultaneously, the internal speech plans are returned to the conceptualizer for monitoring of the degree of success of the conceptualization, formulation and articulation of the message intent, in light of the relevant discourse and encyclopedic knowledge of the speaker.

As this brief description of Levelt's model indicates, his model of production is heavily lexically driven. Lexical selection is considered to drive grammatical encoding. As lemmas are retrieved when their semantic conditions are met in the message, they activate syntactic procedures that correspond to their syntactic specifications. Levelt's model illustrates the putative process in which the grammatical encoder syntacticizes the preverbal message using the syntactic specifications provided in the retrieved lemma in order to derive a surface structure of the message. The surface structure is then processed in the phonological encoder for exact form specifications, which is then sent to the articulator to derive overt speech.

It is the grammatical encoding in this process that requires a focus on syntactic form on the part of the language producer. Although essentially the reverse process is believed to take place in the speech comprehension system for any incoming language input, the additional knowledge source stored in the discourse models and situational and encyclopedic knowledge can often compensate the lack of L2 knowledge in decoding the input data. The grammatical decoding, therefore, may effectively be bypassed in the course of input comprehension, as learners use top-down processes in comprehension. In production, on the other hand, the speaker is responsible for message generation and formulation that requires grammatical encoding. There is much less chance for the speaker to escape syntactic operations in the course of production. It is in this sense that output is said to force the learner to move from “semantic processing prevalent in comprehension to the syntactic processing needed for production.” (Swain and Lapkin 1995)

2）Skill acquisition theories and the development of fluency in production

It is widely accepted that output plays a direct role in enhancing fluency/automatization and in acquiring language skills. De Bot (1996), for example, claims that “the locus of the effect of output must be in the transition of declarative to procedural knowledge” (p529). But scholars remain fundamentally different as to what fluency means. Some view fluency and automatization as the ever more efficient use of rules (cf. esp. Anderson, 1993), while others view fluency and automatization as the ever faster retrieval of instances from memory (cf. esp. Logan 1988, 1992). And theories of fluency development differ accordingly. This section reviews theories of fluency development and how output functions in the skill acquisition.

The psychological mechanisms underlying fluency is an issue of debate for the time being. According to Schmidt (1992), de Bot (1996), Skehan (1999) and DeKeyser (2001), there seem to be three ways of accounting for the development of fluency: accelerating models, instance models, and restructuring models.

The first approach, accelerating models, or rule-based approach, simply suggests that there is a natural sequence in which initial declarative knowledge becomes proceduralized (Anderson, 1989) or automatized, so that essentially similar processes are used, but more quickly and with less need to use mental resources to control them. When this procedure/connection is made repeatedly, the activity becomes automated, and therefore more rapid and more precise. This model is easily illustrated through the acquisition of skills, such as driving a car, where procedures which initially dominate consciousness and consume all available attention, such as changing gear, recede in importance as they become routinized.

In this model, output plays a direct role in enhancing fluency by turning declarative knowledge into procedural knowledge. According to de Bot (1996), output does not play a role in the acquisition of completely new declarative knowledge, because learners can only acquire this type of knowledge by using external input. But output plays a role in skill acquisition because acquisition implies the development from controlled processing to automatic processing, and that fluency on one level allows attentional resources to be spent on higher-level processes.

In contrast, instance-based approaches (Logan 1988; Robinson and Ha, 1993), regard all learning as encoding and storing of instances and performance is retrieval of instances. Essentially, an instance is a representation of co-occurring events; which co-occurrences are remembered is determined by what the learner pays attention to. Thus, fluency or automatization is the product of organizing performance so that it is based not on rules which are applied more quickly, or on rules which are more efficiently organized, but on contextually-coded exemplars, such as a particular example of a past tense form. These exemplars can function as units, which may have been the product of previous rule application now stored in exemplar form, and which require far less processing capacity, since such exemplars are retrieved and used as wholes. In this view, learning is the result of instance creation, and performance (and the ensuing fluency) the result of instance use.

In instance-based model, output and consistent practice serve to increase the likelihood of retrieving an item by strengthening its representation. In other words, output and consistent practice increase the probability of fast retrieval by increasing the number of representations (instances encoded). Fluency is observed in reaction times and is the result of the power law of learning.

Restructuring approaches (McLaughlin, 1987, 1990), regard improved performance as a qualitative change, resulting from restructuring of the underlying mechanisms involved in carrying out the performance. In other words, improved performance is the result of using better algorithms so that performance is better (and differently) organized, for example, sorting out the rule underlying tag forms, such as “isn't it”. The acquisition of cognitive skills/procedures does not develop in a linear fashion, but undergo constant tuning and restructuring. The learner constantly imposes organization and structures on the information that has been acquired. As more learning occurs, internalized, cognitive representations change and are restructured. This restructuring process involves operations that are different from, but complementary to, those involved in gaining automaticity (McLaughlin, 1987). Following this approach one assumes that restructuring, when it occurs, is rapid, and immediately available to sustain improved performance (fluency in this case).

Each of these models has relevance for fluency, but in contrasting ways. The proceduralization model and the restructruring approach (a reorganization of the rule-based analytic system) emphasize the role of rule-based system in developing fluency, while the instance theory stress on the exemplar-based system. There are signs of convergence between these two approaches, though. Palmeri (1997) and Anderson et al. (1997) stress the importance of analogy to examples in the first (declarative) stage of skill acquisition, but they also acknowledge the possibility of item retrieval after the relevant procedural rule has been compiled. Richard (1997) marks a further rapprochement between the two by showing that the learning curves he found for some pseudo-arithmetic tasks can be analyzed as a relative long period of speed-up of an algorithm, followed by a period of increasingly fast item retrieval. This may imply that algorithms are important in the early stages, and that instance retrieval is more important in later stages of fluency development. As was mentioned in section 3.1, production may serve as a link between instance-based system and rule-based system, enabling movement to occur from the rule-based to the memory-based system and vice versa. It is clear that production leads to not only fluency/automatization, but also restructruring of learner language. (see Chapter 4 for detailed information on the concept of “restructuring”).

3）Psycholinguistic mechanisms of language output

With regard to the psycholinguistic mechanisms underlying language output, Swain (1995) points out that production makes the learner move from “semantic processing” prevalent in comprehension to more “syntactic processing” that is necessary for second language development. In other words, output may stimulate learners to move from the semantic, open-ended, non-deterministic, strategic processing prevalent in comprehension to the complete grammatical processing needed for accurate production. Izumi (2003) argues that the processes of grammatical encoding during production and monitoring to check the matching of the communicative intention and the output enable language learners to assess the possibilities and limitations of their interlanguage capability. This may, under certain conditions, serve as an internal priming device for consciousness-raising for the learners, which in turn creates an optimal condition for language learning to take place.

The psycholinguistic mechanisms are summarized in Figure 3.3. According to Swain and Lapkin (1995) and Izumi (2003), what goes on between the original output (Output 1) and its reprocessed form (Output 2) is part of the process of second language learning. When facing problems in their production process, learners have several alternative routes to take depending on the given situation at the time. These alternative routes include receiving external feedback to test the interlanguage hypothesis, consulting a “authoritative” figure for assistance, utilizing shared resources if learners work collaboratively, exploring the internal procedures and internal resources for a possible solution, or paying attention to the subsequent input. In other words, output processing engages important internal procedures such as grammatical encoding and monitoring, which prompts the learners to interact actively with the external environment to find a solution or to explore their internal resources for possible solutions. Output, thus, serves as a useful means to promote the interaction between learner internal factors and environmental factors, or the interaction within the learners themselves for internal metalinguistic reflection. The processes that intervene between the first output and the second output, which are depicted in the squares, are believed to constitute an important part of SLA.Figure 3.3　Output and second language development (from Izumi, 2003, p187)

Thus, according to Swain, output promotes acquisition either through engaging syntactic processing or through automatizing learner's discourse and linguistic knowledge. Output is an active component in the overall SLA process. However, only a limited number of these functions have been empirically tested, and of these, very few have been addressed within a cognitive framework.

4）Robinson's cognition hypothesis

If the role of output is not limited to promoting fluency, one more question may be raised: does output and language production lead to accuracy and complexity? It is still doubtful whether output can also enhance accuracy and complexity of learners' interlanguage. Theoretically, this is possible as long as tasks are designed and sequenced in such a way as to direct learners' processing capacity to grammaticization, and so encourage greater accuracy and complexity of L2 production. In this aspect, Robinson (2001, 2005) proposes Cognition Hypothesis which can predict linguistic and conceptual development during language production. Let's have a brief look at the Cognition Hypothesis.

Robinson makes a distinction between L2 performance and L2 development. L2 performance corresponds to those dimensions of task demands which can be manipulated to stimulate access to an existing L2 knowledge base (such as allowing planning time, see, e.g., Skehan 1999). L2 development corresponds to those dimensions of task demands which can be manipulated to push learners to go beyond this to meet the demands of a task by extending an existing L2 repertoire (such as making increasing demands on the conceptual/linguistic distinctions needed to refer to spatial location, temporality, or causality, see, e.g., von Stutterheim and Klein 1987). This distinction makes it clear that both L2 performance and L2 development are possible during language production. Then, he proposes the Triadic Componential Framework for task classification and design. This framework, as illustrated in Figure 3.4, distinguishes the cognitive demands of pedagogic tasks contributing to differences in their intrinsic complexity, from the learners' perceptions of task difficulty and task conditions. Of these three sets of factors he suggests complexity differentials alone should be the major basis for proactive pedagogic task sequencing in task-based syllabuses.Figure 3.4　A triad of task complexity, task condition and task difficulty factors (based on Robinson, 2005)

According to Robinson, there are two categories of the dimensions of task complexity: resource-directing dimensions, and resource-dispersing dimensions. Increasing task complexity along resource-directing dimensions (e.g. by requiring reasoning, in addition to simple information transmission) makes greater resource demands which can be met by using specific features of the language code. In this case those features might include use of logical connectors (if—then, therefore, because) along with the syntactic permutations necessary to embedding and subordination of clauses. Increasing complexity along these dimensions therefore has the potential to direct learners' attentional and memory resources to the way the L2 structures encode concepts, so leading to interlanguage development. In contrast, complex tasks along resource-dispersing dimensions make greater demands on attention and working memory, but do not direct resources to features of language code that can be used in completing the task.

Robinson (2005) argues that recent studies support three predictions of the Cognition Hypothesis. These three predictions are that1) task complexity along developmental dimensions leads to less fluency, but increased accuracy and complexity of production; 2) task complexity leads to greater amounts of interaction and negotiation for meaning, and subsequent noticing and incorporation of input in learners' production; and 3) IDs in relevant clusters of cognitive abilities increasingly differentiate performance as tasks increase in complexity (p11).

He argues that increasing the functional/cognitive demands of communicative tasks has the potential to affect the way L2 production is syntacticized, i.e., to cause a shift from the pragmatic to syntactic mode, or to push development beyond the basic learner variety. Increasing the complexity of the conceptual and functional demands of tasks is also likely to draw learner attention to the ways in which the L1 and the L2 may differentially grammaticize conceptual notions, and so have positive effects on L2 accuracy of production. This process is similar to the conceptual development in childhood in L1, and the

试读结束[说明：试读内容隐藏了图片]

下载完整电子书