Research Methodology in Second-Language Acquisition
Research Methodology in Second Language Acquisition brings together papers which address serious concerns about Second Language Acquisition (SLA) research. The research methodologies examined, almost always with applications, range from discrete point grammaticality judgments to qualitative studies examining group work during an ESL class. A few chapters introduce new methodologies and most chapters consider ways for improving or refining existing methodologies. While some research methodologies receive little or no attention, the research methodologies examined are discussed in a principled manner. The different chapters explore issues in SLA research methodology, not by telling the reader what should be done, but by showing through the use of actual studies how methodologies can be improved. The introduction and the concluding chapter furnish useful perspectives on the larger issues.
Perhaps the most important issue addressed in this book is whether a single methodology exists for SLA research. The answer is clearly no. Different methodologies possess different strengths and weaknesses. Indeed, Tarone's concluding chapter suggests very practically that researchers use the approach which best suits the questions being explored. No one approach to research fits the different questions researchers seek to answer.
The book consists of three sections: Evaluating Competing Frameworks, Methodologies for Eliciting and Analyzing Language in Context, and Methodologies for Eliciting and Analyzing Sentence-Level Data. These section titles present an overview of the general concerns of the book. Space does not allow an in-depth exploration of the large variety of issues explored in the different chapters. I would like instead to examine a few issues which cut across the chapters.
First, we have the problem of understanding the data. Markee challenges the dominance of nomothetic (assuming a single reality for discovery) at the expense of the hermeneutic (interpretation of the multiple existing realities) scientific tradition. He illustrates the strength of an ethnomethodologic approach using conversation analysis to document a student's learning the word coral in a classroom setting. The strength of this approach lies in the capability to get a better grasp on the complexity of SLA by examining acquisition in a naturalistic setting, here the classroom. [-1-] Continuing in this vein, Douglas and Selinker show how context in Language for Specific Purposes can be explored through triangulation of the data by using various viewpoints to examine interlanguage development, including, unlike the ethnomethodologic approach, those of subject-specific informants, the subjects themselves, as well as language specialists. Cohen and Olshtain explore ways in which data for speech act research are collected and analyzed, pointing out the strengths and weaknesses of various methods used, as well as pointing out the insights gained from involving the subjects in the evaluation. Another possible approach for getting more than one point of view comes through use of collaborative grammaticality judgments, which, as Goss, Zhang, and Lantolf illustrate, result in similar judgments but may provide a better access to what the subjects are thinking about than the think-aloud studies do. Kathleen Bardovi-Harlig argues for the quantification of qualitative data to enable comparison, which would allow researchers to move, as she phrases it, "from anecdote to evidence."
These chapters represent what Tarone, in the concluding chapter, characterizes as a profligate or open-ended type of data gathering. They focus upon the external aspects of language acquisition. In contrast, the conservatives limit their data gathering in order to attempt to gain insight into the internal aspects of language acquisition.
This second group concerns itself with various threats to validity and reliability in attempting to refine the research methodologies used. The final section of the book devotes seven chapters to grammaticality judgments. Munnich, Flynn, and Martohardjono discuss the use of production through elicited imitation, Bley-Vroman and Chaudron point out problems with this type of research, Cowan and Hatasa argue for the use of on-line tasks, and Gass supplies an approach to making grammaticality judgments more reliable through eliminating areas of indeterminate knowledge.
A second issue raised concerns the relationship between SLA theory and Universal Grammar (UG) theory. Many chapters in the book detail experiments which test a UG principle. The problem arises when the data suggest that the interlanguage contradicts principles of UG.
Hagen's chapter questions SLA researchers investing so heavily in UG theory while not considering competing theories. He concludes by arguing that SLA theory construction and testing need not tie itself to UG. In fact, findings that support UG, but are arrived at independently, will be stronger than research motivated solely by UG. Eckman finds contradictions between his subjects' interlanguage and UG, but holds out the possibility that some language not yet discovered may use this pattern. Berent explores research related to the Subset Principle. This principle posits that acquiring a [-2-] language involves constructing hypotheses based on positive evidence, which are used in building the smallest language compatible with these hypotheses. SLA research has found this theory wanting, but Berent, in reviewing existing studies, identifies problems with the validity of task designs and the soundness of theoretical linguistic proposals, and concludes that principled exploration of the Subset Principle in SLA will yield results. However, Lakshmanan and Teranishi's findings contradict the Subset Principle in UG theory. Despite these apparent divergences from UG theory, only Hagen suggests that other available theories be considered or that UG theory may not apply to the interlanguage.
Analysis of the data is another concern. New inferential techniques for data analysis are described by Hagen and Bayley. Hagen uses repeated-measures data to explore the reflexive and non-reflexive pronouns of English and Spanish L1 speakers making grammaticality judgments through selecting the appropriate verb-pronoun pair to complete sentences. Hagen employed Factor Analysis and a repeated-measures ANOVA to analyze the data to determine whether they support the proposed hypotheses. Bayley investigates the variationist debate reviewed in Eckman's introductory chapter. Adopting the quantitative methodology of sociolinguists, he analyzes interview data using VARBRUL2 to determine whether a pattern of acquisition can be discerned. His analysis shows the usefulness of adapting this method of analysis to SLA research.
Aggregate data analysis through various statistical techniques is challenged by Eckman and by Lakshmanan and Teranishi. The fundamental argument is advanced by Eckman, who points out that interlanguage by its nature is individual and needs examination as an individual dynamic to better understand it. By putting the data in aggregates, we loose information about the individuals, perhaps discerning a pattern which upon closer examination we would find no individual exhibiting. This problem has been raised by other authors, notably Runkel (1991) and Cziko (1993), who raise similar concerns about the use of the method of relative frequencies. The method of relative frequencies tells us how a significant number of people behave without demonstrating that any person actually behaved in that way.
The concern for the validity and reliability of the research expressed by the various authors is betrayed by a few proofreading problems. In one case, on page 219, the writer referred to type A when type C was meant. In another case, on page 315, the table states n=24, but only 23 subjects are listed. The first problem caused more difficulty in reading, since I had to check and recheck the item mentioned eight pages previously. These flaws, however, do not condemn the work, and perhaps were only noticeable because of the overall high quality of the editing. [-3-]
A more serious limitation of this collection is the lack of scope; some important research methodologies are not explored. For example, with few exceptions, the research methodologies involve cross-sectional data or test-retest designs. The lack of longitudinal methodologies is one deficiency. A more serious deficiency concerns the lack of ethnolinguistic and case study methodologies. Furthermore, the only naturalistic study concerns a classroom setting. Also problematic is the focus on UG-related theory testing, while no one even mentions, let alone explores, the competition model of psycholinguists Bates and MacWhinney.
A common theme which emerges from these chapters portrays SLA research as still developing, with a need to build cumulatively upon previous research methodologies. This theme gets artful elaboration through the different approaches to solving problems of collecting and analyzing the data. While some approaches point us in the direction of further refinements of statistical analysis, other approaches expand the ways of collecting and examining the data. The chapters collected in this book provide impetus with guidance for the continued principled development of SLA research by building thoughtfully and critically upon existing methods while proposing improvements. Most SLA researchers will find something of value on first and later readings, whether they are concerned with their own methodological problems or with understanding other approaches.
Cziko, G. (1992). Purposeful behavior as the control of perception: Implications for educational research. Educational Researcher, 21, 10-18.
Runkel, P. J. (1990). Casting nets and testing specimens: Two grand methods of psychology. N Y: Praeger.
John M. Graney
University of Illinois, Urbana-Champaign
© Copyright rests with authors. Please cite TESL-EJ appropriately.
Editor's Note: Dashed numbers in square brackets indicate the end of each page in the paginated ASCII version of this article, which is the definitive edition. Please use these page numbers when citing this work.