On the Functions of L2 Speech Production

and Related Cognitive Processes for the Acquisition of L2 Speech Competence

1. Introduction

The learner's oral output plays an important role in his or her acquisition of L2 competence. For instance, it is the basis for corrective feedback, it allows for syntactic processing, and it promotes the automatization of production processes. Moreover, oral output is a precondition for the active testing of hypotheses and the acquisition of discourse competence. Up to now, research on the acquisition of foreign language competence has predominantly investigated the structure and the functions of the input and as a consequence developed mainly input-oriented production models. In contrast, we consider oral output central to L2 acquisition. We are therefore trying to identify further functions of the output and to investigate interdependencies between the learner's linguistic competence and his or her performance. Oral production is the result of various cognitive processes, among them attention, monitoring, and automatization. The aim of our project is to integrate these three factors into a unified model of L2 speech production and acquisition.

2. The research context

a) Speech production models: Levelt's (1989) model of speech production is probably the most influential one in psycholinguistics, but it is limited to monolingual contexts only. Its three main components conceptualizer, formulator, and articulator depend on each other in a unidirectional way. De Bot (1992) has adapted Levelt's model which is based on empirical data to the bilingual competence of L2 learners. Both models have been criticized for their low degree of flexibility. It has also been questioned whether they are able to account for the typical variation in the performance of L2 learners and the characteristics of multilingual speech production processes (cf. Poulisse & Bongaerts 1994; Williams & Hammarberg 1998). In contrast to these two models, we assume that L2 performance as well as related cognitive processes play a major role in the development of L2 competence. In consequence, we do not subordinate output to input but consider it as being equivalent.

b) Attention: Attention is generally regarded as a necessary precondition for successful learning. While research related to the noticing hypothesis (Schmidt 1990; 1995) has investigated the characteristics of attention directed towards the input, attention directed towards the output has received almost no attention yet. In our study, we therefore manipulate the focus of attention towards several aspects of speech production.

c) Monitoring: Monitoring is a special type of attention process which can be observed in any kind of speech performance (both in L1 and L2). Speakers evaluate their output in terms of the variables form, content, and situation. The investigation of monitoring processes in L2 production is likely to yield insights into how learners make use of their linguistic knowledge and which role this (explicit or implicit) knowledge plays for them. Indicators of monitoring processes can be inferred from self-repairs (cf. Kormos 2000) or retrospective data. It is mainly the latter which reveal planning intentions and production processes of L2 speakers.

d) Automatization: Automatized processes in L2 acquisition require no or only little attention. Hence, one of the most typical features of automatization is that cognitive tasks demanding a large amount of attention can be processed at the same time as automatized processes take place. Automatization, therefore, is a necessary precondition for the enhancement of fluency.

e) Preliminary work: A cross-sectional study at the University of Bielefeld entitled "Features of L1 vs. L2 speech production under special consideration of fluency" has already analysed the oral speech production of 13 GFL (German as a foreign language) learners. Their mother tongue was Hungarian or Spanish respectively. In a subsequent five months longitudinal study, two of these learners were confronted with a number of different speech production tasks. The results of this qualitative study indicate that fluency acquired too early can impoverish speech correctness and thus lead to a stagnation in the development of L2 competence. Furthermore, the data illustrate the significance of interaction and attention for L2 acquisition. As for the research methods, above all introspection and retrospection were highly informative in regard to the learner's hypotheses, his or her subjective theories and the cognitive processes involved in L2 production.

3. Research interests and aims

The major aim of our project is to develop an L2 specific speech production model. For this purpose, we systematically manipulate the learner's attention within a 3 x 3 design. While the first factor directs the learner's attention towards his or her oral output (focus on form, focus on meaning, neutral focus), the second factor concerns the social situation (interaction with a higher status native speaker, interaction with a peer native speaker, learner-learner interaction). Monitoring and automatization are analysed with respect to these nine attention conditions.

Within this framework, we examine to what degree monitoring processes are related to the focus of attention. This implies, that first of all, reliable measures for monitoring processes in the oral output are identified. Afterwards, we take a look at the various linguistic dimensions of speech production that monitoring processes operate on (form, content, situation). Furthermore, one may ask whether monitoring is a variable process and whether, in the case of low monitoring, attentional resources are available for different cognitive tasks, e.g. for the planning of speech sequences.

On the other hand, the role of automatization processes in the acquisition of L2 competence has to be investigated in detail. Again, one of the necessary preconditions is the identification of reliable measures for automatization in the oral output of L2 learners. Recent research in cognitive psychology may help to accomplish this task. Furthermore, it is of special interest to see whether, in the course of L2 acquisition, there is a typical chronology for the acquisition of automatization across various linguistic dimensions. Also, we will investigate how the acquisition of automatisms - which is a precondition for fluency - can be guided by manipulating the focus of attention and how fossilizations (i.e. the automatization of inadequate lexical, idiomatic or syntactic structures) can be avoided.

The investigation of monitoring and automatization is closely interrelated insofar as each of the two cognitive factors is characterized by attentional processes or their absence respectively. Consequently, our aim is to integrate the detailed analyses within a strictly interdisciplinary research framework. SLA research may contribute to this goal to the same degree as (psycho-) linguistics or psychology.

4. Methods

The speech production of L2 learners is such a complex object of research that the application of a multimethod research design seems inevitable. For this reason, the speech production of 16 learners is elicited and analysed by a number of different methods (triangulation; cf. Aguado & Riemer 2000). In addition to this primary data, secondary data such as the learner's written performance, his or her individual learning biography, his or her self-assessments or retrospection are of interest. In order to guarantee the compatibility of the various methods and to asses their efficiency, each method is evaluated by means of a tencriteria catalogue.

Participants: 16 foreign GFL students at the University of Bielefeld with different mother tongues. They are all attending an eleven-month language course, at the end of which they will take the DSH (an exam permitting them to begin with their studies at a German university).

Data collection: The collection and analysis of the primary data consists of two major parts. Within a cross-sectional study, how the manipulation of the attentional focus involves differences in the oral output of L2 learners regarding automatization and monitoring processes, is investigated. Based on these findings, changes in automatization and monitoring processes in the course of L2 acquisition can be identified within the framework of a longitudinal study.In each of the nine sessions of the cross-sectional study, participants solve two speech production tasks in one focus condition. In the first task, participants first read a text and then reproduce it. The second task consists of the description of a cartoon. Each session takes approximately 60 minutes. Immediately after each task, a short retrospective phase is conducted during which the participants describe the difficulties they encountered in solving the task. All speech production is recorded by video, audio, and DAT.Within the longitudinal study, a standardized interview technique is employed to account for changes in automatization and monitoring processes in the course of L2 acquisition. Each of the three sessions lasts 20- 60 minutes (at the beginning, in the middle, and at the end of the study). After a short warm-up, the interviewer asks the participants increasingly difficult questions, thus leading them to the limits of their proficiency. These data allow for an analysis of progress in terms of fluency (as a consequence of automatization in phonetics, prosody, syntax, and lexis), pragmatic adequacy, and correctness. These categories also permit an assessment of the L2 learner proficiency.

In order to collect secondary data, the following methods are being used:

1. A language proficiency test of the University of Bielefeld serves as a preliminary means of assessing the learners' linguistic competence (listening comprehension, reading comprehension, grammatical transformations, lexical tasks, free text production).

2. A questionnaire elicits biographical data from the participants, aspects of their learning biography, and individual learning strategies and difficulties. Furthermore, participants are asked for self-assessments in regard to their linguistic competence in grammar, lexis, phonetics, and pragmatics.

3. A specially designed test (SW-Test) accounts for linguistic knowledge in the field of morphology, syntax, and lexis/semantics. For this purpose, various types of tasks are used. Supplementary data consist of explanations, motivations or intentions, which the participants uttered during a think-aloud task.

Both, the questionnaire and the SW-Test are administered three times: at the beginning, in the middle, and at the end of the longitudinal study, which lasts nine months in all.

Data analysis: Due to the multimethod research design, a large amount of quantitative and qualitative data must be integrated into a holistic interpretation. This may serve as the basis for the major aim of our research project: the development of an L2 specific speech production model. For this purpose, the following analyses need to be performed:

1. First of all, the digital audio data must be transcribed with the help of special software (Transcriber).

2. The resulting transcripts are to be analysed by means of L2 specific discourse analysis (cf. Henrici 1995). Then the data can be categorized according to correctness and situational adequacy.

3. The written data (secondary data) allow for the categorization of errors and mistakes regarding morphosyntax, syntax, and lexis.

4. Temporal analyses of the digital audio data, e.g. with Transcriber or Praat, may yield insights into the development of fluency. Moreover, they may help to identify automatization (e.g. speed of production, articulation speed, lack of pauses within automatized sequences) as well as monitoring processes (e.g. pauses, self-repairs etc.).

5. The analysis of the number and quality of semantic units (propositions) reveals semantic aspects of oral speech production.

6. Finally, the retrospective data can be categorized with respect to the difficulties, which participants encountered in solving the speech production tasks.

5.References

Aguado, Karin & Riemer, Claudia (2001): Triangulation: Chancen und Grenzen mehrmethodischer empirischer Forschung, in: K. Aguado & C. Riemer (Hg.): Wege und Ziele. Zur Theorie, Empirie und Praxis des Deutschen als Fremdsprache (und anderer Fremdsprachen) Festschrift für Gert Henrici zum 60. Geburtstag (S. 245-257). Baltmannsweiler: Schneider.

de Bot, Kees (1992) : A bilingual production model: Levelt's 'speaking' model adapted. Applied Linguistics 13, 1-24.

Henrici, Gert (1995): Spracherwerb durch Interaktion? Eine Einführung in die fremdsprachenerwerbsspezifische Diskursanalyse. Baltmannsweiler: Schneider.

Kormos, Judit (1999): Monitoring and self-repair in L2. Language Learning 49, 2, 303-342.

Levelt, Willem J.M. (1989): Speaking: From Intention to Articulation. Cambridge, MA: MIT Press.

Poulisse, Nanda & Bongaerts, Theo (1994): First language use in second language production. Applied Linguistics 15, 1, 36-57.

Schmidt, Richard (1990): The role of consciousness in second language learning. Applied Linguistics 11, 129-158.

Schmidt, Richard (1995) (Hg.): Attention and awareness in foreign language learning. University of Hawai'i at Manoa: Second Language Teaching & Curriculum Center, 1-63.

Williams, Sarah & Hammarberg, Björn (1998): Language switches in L3 production: implications for a polyglot speaking model. Applied Linguistics 19, 3, 295-333.