THE PHILOSOPHY OF SCIENCE AND THE SOCIAL-COGNITIVE DICHOTOMY IN RESEARCH IN LANGUAGE LEARNING AND TEACHING

Robert DeKeyser

University of Maryland

Ever since the cognitive-versus-sociocultural debate that took place in various journals and at several conferences in the 1990s, many researchers in L2 learning and teaching are probably convinced that a wide gap between the two camps is unavoidable. Yet, both Jim Lantolf and I started off our individual contributions at the AAAL 2013 colloquium by saying that there does not need to be a gap. By the standards of the 1990s, we were off to a good start!

Why does there not need to be a wide gap? Why is it even surprising that we have a (perceived) gap? Obviously, most language learning takes place in social and cultural contexts that play a role in the learning process (and all language learning does, if you call the psycholinguistics laboratory a sociocultural context, too, albeit a highly marked one). Equally obviously, all social and cultural learning is at least in part a cognitive process: We acquire social and cultural knowledge, and knowledge and cognition are the same thing. Note that they are even etymologically related: The -gn- in cognition and the kn- in knowledge go back to the same Indo-European root (e.g., see Watkins, Reference Watkins2000, pp. 32–33).

If that is all true, then why is there a widely perceived gap? The core problem, in my opinion, is that a number of dichotomies that have little to do with one another in principle do largely coincide with the cognitive-social dichotomy in practice; this is apparent, for instance, in the way language learning research is currently practiced: Research can be quantitative versus qualitative, hypothesis testing versus hypothesis generating, and absolute versus relativist, and it can use truth versus esthetics as evaluation criteria. Equating all of these dichotomies amounts to confusing material objects (cognitive or social phenomena) with formal objects (aspects of the phenomena most amenable to certain disciplines, methodologies, approaches, etc.; e.g., see Maritain, Reference Maritain2005, p. 67).

In principle, the quantitative-qualitative distinction does not belong here at all. Counterexamples abound of the cognitive equals quantitative and social equals qualitative equations. Prominent examples of cognitive research that is not quantitative in nature are Piaget’s work in developmental psychology, Chomsky’s approach to language (which is not quantitative, and not even experimental), and much of the early work on first language (L1) acquisition by researchers such as Grégoire (1937–1947), Leopold (1939–1949), Lewis (Reference Lewis1936), or Weir (Reference Weir1962), which was, again, mostly not quantitative and not experimental in nature. Conversely, we all know of research in the social and cultural domains that is heavily quantitative in nature, such as the research on motivation for L2 learning in the social psychology tradition (e.g., Gardner, Reference Gardner1985; Lambert, Reference Lambert1972; MacIntyre, Reference MacIntyre and Robinson2002) or Gelfand’s work (e.g., Gelfand et al., Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim and Yamaguchi2011) in social psychology and cultural anthropology that quantifies essential features of a culture.

More inherent is the link between the quantitative-qualitative dichotomy and the distinction between hypothesis generating and hypothesis testing. Quantitative and qualitative approaches do not simply mean presence or absence of lots of numbers. Although the two have been defined in many different ways, for me the most useful way of thinking about them is that quantitative research stresses reliability over validity and control over context, the opposite being true for qualitative research. As all four of these characteristics are desirable, both approaches have advantages. However, the observation that an emphasis on validity and context is better suited for hypothesis generating and that an emphasis on reliability and control is better for hypothesis testing is inescapable. Both are important, of course; the inductive-deductive cycle of hypothesis generating and testing is what allows for theory building and falsifying. Be that as it may, science strives for generalization (not even the biggest accumulation of facts amounts to science), and that requires hypothesis testing. As most sociocultural work is exclusively qualitative and, hence, does not allow for hypothesis testing, the association between the sociocultural and the qualitative makes a research endeavor that is limited to the sociocultural inadequate from the point of view of the philosophy of science. This is not to say, of course, that a case study cannot be useful. It can be very important to help generate hypotheses for hypothesis-testing research, and it can also serve to refute some hypotheses by providing a counterexample. Much of our research, however, is about how strong (or how likely) the association between some variables is, and the study of a single case cannot do much to provide evidence for the correlation of variables over cases or the strength of correlation. (Of course I am not merely referring to research involving correlation coefficients here; comparing group averages for a given variable is also about establishing a relationship between variables, one variable being group membership in this case.)

As previously pointed out, however, the association between the sociocultural approach and qualitative methodology is a mere fact, not a necessity. I see no reason why sociocultural research could not move from descriptive to explanatory to predictive, as long as the social dynamics it takes as its material object are not confounded with relativism to the point of rejecting the ideas of hypothesis testing and falsification of theory. Without such a process, there is no criterion for selecting among theories (beyond their esthetic or political appeal), no generalizability, and no science.

Meanwhile, back in the real world of language learning and teaching, addressing the content questions that really matter from a theoretical and/or practice point of view is more important than philosophizing about methodologies. How does L2 fluency develop? How does the cognitive basis of proficiency change over time? How do different kinds of practice at different stages of development change overall proficiency and specific skills? These are examples of questions to which all teachers and learners can relate, even if they do not use the same jargon.

To provide answers to such questions, we need mixed-methods research of a longitudinal nature. Mixed-methods research can avoid the poor articulation between the quantitative and the qualitative that happens when they are largely separate bodies of literature; it is much better if one can have a descriptive-correlational-experimental loop (Rosenshine & Furst, Reference Rosenshine, Furst and Travers1973) within a large study and not just in the accumulating body of literature. Longitudinal research is ideal for answering questions that address long-term development, as most of the questions in SLA do. Unfortunately, although both longitudinal research and mixed-methods research are already fairly rare in our field, the combination of the two is largely unheard of. There are practical reasons for this, of course (e.g., the finances needed and the risk of unacceptable levels of participant attrition). Other reasons, however, include the sociology of academia (individuals are reluctant to engage in endeavors that will not bear fruit until long after they should have received tenure or promotion) and probably also the fact that such research would not only combine qualitative and quantitative methodologies, which few academics are equally comfortable with, but also require social and cultural contextualization of the psycholinguistic findings as well as cognitive interpretation of the social and cultural dynamics, two things that even fewer academics are used to.

While we wait for a new generation of more truly interdisciplinary teams of researchers, the best we can do is to encourage the descriptive-correlational-experimental loop by letting various approaches and methodologies inspire one another, and ideally feed into one another, as the loop metaphor implies. Even if we manage to pull off a large-scale longitudinal mixed-methods study, the fact that answers to scientific questions come from bodies of literature and not from individual studies should always be in the backs of our minds. Therefore, the need to read and consult widely and to contribute to the loop (or spiral?) will be with us for a long time to come.

Ultimately, the ability of individual studies to contribute to this cumulative body of work will be best served if we all strive for context and control, for reliability and validity, realizing that sometimes a trade-off is unavoidable. There is no such trade-off between the social and the cognitive; after all, Vygotsky himself emphasized the social nature of cognition. We must all strive for generalizability of our findings, though, whether we are primarily cognitively or socioculturally oriented, and whether we try to improve generalizability by building more complex designs (e.g., aptitude-treatment interaction studies instead of mere method comparisons) or by designing simpler studies while keeping in mind their place in the descriptive-correlational-experimental loop. There is much value to many things beyond science (visual arts, film, literature, literary criticism, journalism, etc.), but without a relentless pursuit of the generalizability of knowledge, there is no (social) science.

A BRIDGE NOT NEEDED

EPISTEMOLOGICAL REMARKS ON A SOCIAL-COGNITIVE GAP IN THE STUDY OF SECOND LANGUAGE LEARNING AND TEACHING

Jan H. Hulstijn

University of Amsterdam

As an observer of the discussions in the 2013 AAAL colloquium, it seemed to me that there appeared to be agreement on two points. First, there does not exist a gap between social and cognitive approaches to L2 learning and teaching insofar as language learning is inherently social and, at the same time, all learning is cognitive by definition. There was, however, no agreement on the weight to be attributed to the social aspect of language learning. For instance, for Lantolf, who adopts a Vygotskian approach, the psyche is inherently social, and there is therefore no opposition between nature and nurture. Culture creates special forms of behavior, it modifies the activity of mental functions, and it constructs new superstructures in the developing system of human behavior. For DeKeyser, however, this claim needs to be tested empirically (it must be falsified or supported). Thus, whereas Lantolf emphasized the ontological status of Vygotsky’s theory, DeKeyser raised issues of an epistemological nature: How can we obtain supporting or conflicting evidence for Vygotsky’s theory, or indeed any theory? I return to this point later in this piece.

Second, speakers appeared to agree on the fact that scientific knowledge and the interpretation of the findings of empirical research is collective in the sense that interpretations are shared by scholars united in what might be called a theory or a school—or what Fleck, in an important but relatively unknown book, called a thought collective, and which, some 30 years later, Kuhn called a paradigm (Fleck, 1935/1979; Kuhn, 1962/1996). Whereas Young, in his contribution to the present article, elaborates on the social-cultural context of scientific knowledge in schools (or thought collectives), here I limit myself to proposing some hermeneutic questions with which one may try to diagnose the nature of the gap, real or apparent, between scholars defending different theories. Do their theories belong to different, incommensurable schools, and, if so, what is the nature of the incommensurability? There may be several gaps in our field, of which some might be bridged, whereas others cannot. An unbridgeable gap clearly exists between relativists and critical rationalists. As DeKeyser put it, does an acknowledgement of the social-cultural context of scientific knowledge stand in the way of hypothesis testing, seeking generalizable knowledge, and evaluating theories? DeKeyser pointed out that there is a difference between acknowledging that all scientific knowledge is embedded in a social-cultural context and claiming that all scientific knowledge is relative; a scholar can agree with the first statement but assume at the same time that there is a best truth that we can find at this point and that it is the scholar’s task to uncover that truth.

Although I place myself in the camp of Popperian critical rationalism (Hulstijn, Reference Hulstijn2013; Jordan, Reference Jordan2004), the AAAL colloquium left me in a state of confusion with respect to the question of how to conceptualize differences among applied linguists and SLA researchers in terms of their theories and their epistemological stances. To help myself, and hopefully to help others too, I have listed in the subsequent sections a number of considerations and evaluation criteria instrumental in making sense of differences in theories and their epistemological stances. Though some may consider these criteria biased in favor of critical rationalism, I offer them in the spirit of promoting better understanding across epistemological lines.

GUIDELINES FOR DIAGNOSING DIFFERENCES BETWEEN THEORIES

In trying to make sense of what appear as different theories, approaches, frameworks, or models espoused by individual scholars, it might be useful to ask the following questions.

ADDITIONAL CRITERIA

Given two theories that appear to address the same question and after having answered the four questions just discussed, one might subsequently profit from the following criteria, proposed by Kuhn (Reference Kuhn1977), who thereby demonstrated that he was not an extreme relativist, as some of his critics have argued (see Jordan, Reference Jordan2004, p. 51). I render the criteria in my own words:

• • Coherence. How coherent are the theories? Are their constructs clearly defined, and is their formulation consistent?

• • Testability. To what extent are the theories testable? What is the empirical evidence for the theories?

• • Scope. What is the scope of the theories? That is, which and how many phenomena can they account for (beyond the question that they both address)?

• • Fruitfulness. How (potentially) fruitful are the theories? Do they open avenues for new insights and new research?

• • Simplicity. How simple are the theories? If the two theories score equally well on the other four criteria, can we then, for reasons of economy, select the one with fewer constructs?

I acknowledge that, for extreme relativists (including postmodernists and constructivists; see Jordan, Reference Jordan2004, Chapter 3), these five criteria for the evaluation of competing theories are unacceptable because of their intellectual origins in Popperian critical rationalism. In their view, critical rationalism should not be applied—at least not wholesale—to applied linguistics in general or to sociocultural issues in L2 learning and teaching and language education in particular.

CONCLUDING REMARK

This brings me back to what I see as an almost impossible job of playing the role of a commentator in a more neutral way, with the intention to seek bridges, while, at the same time, embracing Popper’s (Reference Popper1959) philosophy of critical rationalism. When I compare the debate and personal communications during and after the colloquium with the relativist-rationalist debate in the 1990s (e.g., with Block, Reference Block1996, in one camp, and Gregg, Long, Jordan, & Beretta, Reference Gregg, Long, Jordan and Beretta1997, in the other—for more references to this debate, see Jordan, Reference Jordan2004, pp. 1–3), it occurs to me that there are two differences between then and now that give reason for optimism. First, as I tried to argue with my diagnostic questions, there is no single social-cognitive gap in L2 learning and teaching research: Differences in stance may be a matter of epistemology (extreme relativists vs. extreme rationalists), but they may alternatively constitute a matter of ontology (the part of the theory that is proposed as untestable). Second, there appears to be a willingness to listen to voices from the other side and to enter into a dialogue not characterized by acrimony.

Part 2: Data and Research Methods

EXPLORING QUESTIONS OF BALANCE IN INTERACTION RESEARCH

Alison Mackey

Georgetown University and Lancaster University

In the 2013 AAAL colloquium and in my contribution to this article, I was asked to address the following questions: What are the methods of inquiry that you prefer in your own work, what do you see as their weaknesses or limitations, and what attracts you in other methodologies? I have chosen to address these issues by discussing work within the interaction approach to SLA. In particular, I focus on how that approach is currently evolving to include a social dimension and how typical methods of inquiry associated with it are expanding in parallel. I have titled my contribution “Exploring Questions of Balance” for reasons I hope will soon become obvious.

In the initial version of the interaction hypothesis, meaningful interaction, corrective feedback, and associated interactional features were thought to be associated with attention, in a process involving intake and driving L2 learning. However, in the original formulation of the hypothesis and until recently, there were relatively few concrete statements or claims in relation to the social factors that underlie interaction. In the last few years, a reenvisioned and current version of the interaction approach has begun to take off (see, e.g., Mackey, Reference Mackey2012). In general, interaction researchers are now taking a more cyclical view, as DeKeyser has noted in his contribution to this article. Qualitative research often provides insights into social factors such as those mentioned by Young in his contribution to this article, and it is then sometimes followed up by quantitative research. This is essentially the same process that occurs when controlled and balanced laboratory research uncovers constructs that can then be explored in classrooms and possibly applied (with caution, as Lightbown, Reference Lightbown2000, has reminded us) in authentic instructed settings. Keeping this cycle in mind is one of the issues of balance. We need to carefully weigh the cognitive, the social, the laboratory, and the classroom, bearing in mind the goals of the research. It is also important to consider pace, because research needs to progress at a rate at which the expansion of inquiry evolves at a comfortable speed. In other words, we need to recognize the value of different approaches, both in their own right and also for what they can provide. It is not practical and should not be necessary for those interactionists beginning to consider social variables to completely shift their methodological tools and paradigms. Although progress is inevitable as we work to continue to address interesting questions and to elicit valid data, these developments should build on a strong foundation, and we must recognize that the process of evolution is important.

The progression to include social variables can be seen in other work related to interaction research. For example, in relation to Robinson’s (Reference Robinson2001) cognition hypothesis and task complexity, recent work reported by Révész and Gilabert (Reference Révész and Gilabert2013) described methodological advances in task research illustrating the same trend that we are also privy to in the interaction and feedback area—namely, a primarily cognitive orientation beginning to be influenced by social concerns. Révész and Gilabert, for example, looked at learners’ perceptions of difficulty and mental effort, supplementing external theories and claims about difficulty by triangulating data from the learners’ perspectives. It is a short next step for research like this to examine a range of factors that may impact how learners respond to these sorts of questions, and it would seem to be a logical next step from looking at what learners believe about difficulty to asking what contributes to those beliefs.

The realization that considering social and contextual factors provides interesting primary and supplementary data can also be seen in work that looks outside the head of the individual and unpacks what a “task” means to an individual learner or pair. For example, Young’s (Reference Young2009) work recognizes that what he calls discursive practices—essentially communicative tasks—have a history, and he observes that we need to understand what learners do in any one task by also looking at the various aspects of their history with that task and what it means for them.

In other words, as the interaction hypothesis and task-based approaches to learning continue to evolve, it seems logical to me that those taking a more cognitive approach and those taking a more social approach will benefit from working together. This is another one of the questions of balance I want to raise. We need to ask one another, and ourselves, how we can benefit from combining our perspectives. It is clear how fruitful cognitive research and social research are on their own, as other contributions to this article demonstrate. I suggest that we need to ask questions like the following: What can we gain? How can cognitively oriented research most productively include socially oriented methodologies and vice versa? What are the challenges in doing this? How do we go about creating research questions that do not all suppose a priori constructs? And how do we adopt new methods, adapt current methods, and combine methods? I suggest that, to do this most effectively, collaboration is an essential requirement.

In a study I carried out with Jenefer Philp, published in 2010 in a book edited by Rob Batstone, we saw how relationships among learners impacted what they were willing and able to listen to and attend to in interaction and how this impacted what they produced. Because input and output were clearly equated with learning for us as interaction researchers, we realized that we needed to recognize what creates conditions for what we think of as successful input and output. This led us to think about methodologies for investigating social concerns.

What the Philp and Mackey (Reference Philp, Mackey and Batstone2010) study led me to believe is that my own perspective has evolved and now falls somewhere between quantitative and qualitative and between a primarily cognitive and a primarily social approach. This led me to ask how researchers from each perspective do what the other party is trained to do when (a) we do not have the skill set in methodology and (b) we are not fully confident in the framework. These issues are taken up much more eloquently by Martha Bigelow in her contribution to this article. As an example of the need for caution when examining questions of balance, our field does sometimes see criticisms of mixed methods as well as objections when someone like me, who has traditionally taken a primarily cognitive approach, takes a mixed-methods approach. It is obviously a legitimate concern to suggest that tacking a minor case study of one or two learners onto the end of a primarily cognitive, quantitative study is not sufficient to allow us to think of our approach as authentically socially informed. I do not intend this to be an argument against mixed-methods research, although I do suggest that it might be short sighted for social researchers to call for cognitive researchers to examine social factors without acknowledging that there may be benefits to their own research of taking cognitive factors into perspective. The problem is, once again, one of training, of time, and of expertise. Again, I believe one way to address this is through collaboration and teamwork.

Simply associating cognitive with quantitative and social with qualitative is overly simplistic (a point also made in a much clearer way by Talmy in his contribution to this article). Some modes of inquiry are associated with both and some with only one. Some of the constructs that we may consider socially informed, or at least useful for broadening the horizons of traditionally cognitive work—for example, motivation, willingness to communicate, or introspective measures—are psychological constructs and methods that have been included in research on language learning and teaching since the inception of the field. We need to recognize that, just as we read in research methods books, there is no clear dichotomy between quantitative and qualitative, and our familiarity with paradigms similarly operates on a continuum.

I also believe that being open to new methodologies for inquiry, as I optimistically think we have traditionally tended to be in L2 learning and teaching research, may guard us from entrenchment in theoretical approaches and from oppositional positions. Indeed, data obtained through stimulated recall protocols—elicitation techniques that were originally employed in interaction work simply to shed light on learners’ cognitive processes (Gass & Mackey, Reference Gass and Mackey2000)—were what eventually convinced me that social factors are an important force in cognitive-interaction research.

In summary, then, it is my belief that we need, as Rod Ellis (Reference Ellis and Batstone2010) has put it, to go beyond the psycholinguistic rationale in the interaction approach. However—and this is again related to balance—let’s not throw the baby out with the bathwater. I vote for augmenting and developing the interaction hypothesis and working together with those who take a primarily social approach. Together, I think we can achieve the best outcome. By this, I mean research partnerships of paradigms and methods. Social factors underlie the nature of learners’ participation in interaction and, therefore, will logically impact learning opportunities through interaction.

TOWARD AN INTERPRETIVIST TURN IN L2 STUDIES

QUESTION 1. ARE WE BRIDGING DIVISIONS OR (RE)CREATING THEM?

This question concerns the premise of the colloquium itself, specifically the binarisms that it denotes and connotes—cognitive-social, quantitative-qualitative, experimental-naturalistic . . . and why stop there: neutral-political, conservative-progressive, right-wrong. It is hardly an original insight to note that binarisms like these are reductive and fail to capture the kinds of complex inquiries often undertaken in their name. It is worth asking if we might be reifying and reproducing the very divisions the colloquium is predicated on bridging. Moreover, are the distinctions between them as stark as we make them out to be (Allwood, Reference Allwood2012)? Who profits from them? Do we need to bridge divides or dissolve them by interrogating (how we participate in) their construction and by committing to metatheoretical dialogue with colleagues on other coastlines?

I would like to extend this question to the participation structure that was adopted for the colloquium in Dallas: that of debate. Debates cannot be construed as entering into dialogue. Rather, debates yield the sorts of displacive discourse that Julian Edge (Reference Edge2004) has written of (also see Tompkins, Reference Tompkins1988), the stuff of the so-called paradigm wars of the 1970s and 1980s, and, closer to home, in L2 studies, of the 1990s. Have those zero-sum, scorched-earth disputes (of many examples, see Beretta, Crookes, Gregg, & Long’s, Reference Beretta, Crookes, Gregg and Long1994, “response” to van Lier, Reference van Lier1994) taught us nothing?

QUESTION 2. QUANTITATIVE : COGNITIVE :: QUALITATIVE : SOCIAL?

Relatedly, in arguing for an enrichment of research that engages with diverse theoretical orientations, we could also ask if we are working from a premise in which cognitive is mapped to quantitative and social to qualitative. Obviously, there is substantial quantitative research that adopts an explicitly social orientation, just as there is qualitative research that examines cognition. Why not historicize and contextualize our discussion here in those diverse and robust literatures?

QUESTION 3. SAME TERMS, DIFFERENT MEANINGS?

And this brings to mind another question: Do we share similar conceptual vocabularies? Though the terms may be the same—cognition, social, research, learning, bilingual—even the putative learning object, the L2, can mean very different things to those working from, say, positions informed by cognitive psychology or linguistic anthropology, sociology or social psychology, and discursive psychology or linguistics. This is as it should be for a field as increasingly transdisciplinary as L2 studies; nonetheless, it is worth asking if we are playing the same language-game (Wittgenstein, 1953/2001) and, if we can acknowledge that we are not, then asking if there is some way we might contend with that.

QUESTION 4. METHODOLOGY OR METHODOLATRY?

Finally, I was asked to focus my discussion on research methods, specifically, “the methods of inquiry that [I] prefer in [my] own work, . . . their . . . limitations, and what attracts [me to] other methodologies.” But I cannot do this absent an actual study and research purpose. I have conducted critical discourse analyses, critical ethnography, and interview studies, but, in doing so, I have used methods not necessarily associated with these traditions, not because I preferred these methods in the abstract but because they were best suited for the particular research project and research questions I had formulated. So I wonder, is it in our interest to discuss method in the abstract? Might a preference for quantitative methods, or qualitative methods, or, more likely for the purposes of “bridging gaps,” mixed-methods research be putting the cart before the horse? What if we turn our attention to the particularities of our research projects, research questions, and theories and then, in the context of such particularity, formulate our methodology and, in so doing, occupy ourselves not with “methodolatry” (Janesick, Reference Janesick, Denzin and Lincoln1994) but with epistemology and ontology (Pascale, Reference Pascale2011)? Because it is in the areas of epistemology and ontology—not method—that I believe our attention ought to be. And this leads me to the main part of this contribution.

AN INTERPRETIVIST TURN IN L2 INQUIRY

The engagement with ontology and epistemology that I am talking about in terms of method might be called—in its strong version, at least—an interpretivist turn and—in a less strong version, perhaps—a reflexive turn. Regardless of label, such a turn would be suitably rigorous, reflexive, and responsive to diverse theoretical positions and conceptions of language and learning, cognition, the social, and so on (see Sfard, Reference Sfard1998). It would embrace methodological and paradigmatic pluralism and hybridity, diverse “styles of inquiry” (Pascale, Reference Pascale2011, p. 141), similar to Howe’s (Reference Howe2003, Reference Howe2004, Reference Howe2012) notion of mixed-methods interpretivism, a metaparadigmatic move that, when taken, immediately recognizes that paradigms are themselves interpretive frameworks and, thus, that sees all social science research—quantitative, qualitative, and mixed-methods research—for what it is: the outcome of the interpretive activity of particular people working from particular disciplinary and theoretical orientations in particular social contexts and historical moments (Giddens, Reference Giddens1976).

From such a perspective, differing paradigmatic positions and theoretical orientations come to offer different rather than competing perspectives on an object of study—different rather than competing conceptions of what that object might be. Those differences can then be engaged at metaparadigmatic, metamethodological, and metatheoretical levels in ways that reflexively reveal the contours and boundaries of knowledge production and that demonstrate how our conceptions about and our study of a particular phenomenon are inevitably perspectival, as we ourselves as researchers are inevitably implicated in how we come to know and speak of what we study. In short, an interpretivist turn would compel us to avoid the “god trick” that Haraway (Reference Haraway1988) has described. It would oblige us to acknowledge that research is the outcome of our participation in particular social practices, practices that are as diverse as they are generically identifiable, disciplinarily and theoretically informed, historically specific, and paradigmatically sited.

Of the potential ways forward in such an interpretivist turn, one may be with Howe’s mixed-methods interpretivism (mentioned previously). Another could be Alvesson and Skoldberg’s (Reference Alvesson and Skoldberg2009) reflexive methodology, which I briefly sketch next.

REFLEXIVE METHODOLOGY FOR AN INTERPRETIVIST TURN

Reflexive methodology is characterized by careful, rigorous interpretation and a sustained, thoroughgoing reflexivity. Interpretation “implies that all references . . . to empirical data are the results of interpretation. Thus the idea that measurements, observations, the statements of interview subjects . . . the study of . . . statistics or archival data [in some way reflect or mirror reality] is rejected on principle” (Alvesson & Skoldberg, Reference Alvesson and Skoldberg2009, p. 9). Methods are selected not in the abstract but always in relation to specific research problems and research objects (p. 8). This is not to say that inquiry proceeds without criteria for high quality: Rigor counts, certainly, but there is a shift from searching for “truth” to developing well-warranted, defensible claims based on empirical data, which at the same time are acknowledged to be the outcome of the researcher’s analytic activity.

The possibilities of, and dimensions for, high-quality research are further enhanced by reflexivity. But this form of reflexivity does not remain at the level of method; it is “the launching of a critical self-exploration of one’s own interpretations of empirical material” (Alvesson & Skoldberg, Reference Alvesson and Skoldberg2009, p. 9). It is a “problematizing practice” (Pennycook, Reference Pennycook2001, pp. 41–45) that includes consideration of the inherent value-laden-ness of research and of representational politics (Giroux, Reference Giroux1992). These interpretations of the interpretation can then be considered for their epistemological and ontological assumptions, for what they offer, for what they do not offer, and for points of contact and commonalities as well as divergences and disjunctures between alternative interpretive frames. With such reflexivity, “the center of gravity is shifted from the handling of empirical material towards . . . a consideration of the perceptual, cognitive, theoretical, linguistic, (inter)textual, political and cultural circumstances that form the backdrop to—as well as [infuse]—the interpretations” (Alvesson & Skoldberg, Reference Alvesson and Skoldberg2009, p. 9; also see Giddens, Reference Giddens1976; Pascale, Reference Pascale2011).

METHODS AND DATA

RESEARCH IN L2 LEARNING AND TEACHING AS A COLLECTIVE ACTIVITY

In conceiving the 2013 AAAL colloquium, Hulstijn and I were inspired by an idea, originally proposed by Ludwik Fleck (see Sady, Reference Sady and Zalta2012), that what members of a research community know about their field is, in fact, a collective social activity. If we extend Fleck’s (1935/1979) ideas to L2 studies, it becomes clear that, when individuals do research in language learning and teaching, when instructors teach SLA, and when teachers relate research to the classroom, they are part of a community of persons mutually exchanging ideas and maintaining intellectual interaction, a community Fleck called a thought collective. Our understandings of the diverse phenomena of L2 learning and teaching is a product of that collective activity—a thought style—within the thought collective, or a “picture, which is visible only to anybody who takes part in this social activity, or a thought which is also clear to members of the collective only” (Cohen & Schnelle, Reference Cohen and Schnelle1985, p. 77).

In fearing that research in L2 learning and teaching runs the risk of disintegrating into two irreconcilable approaches to language learning and language use, Hulstijn and I were echoing another of Fleck’s (1935/1979) notions: the incommensurability of thought styles, a notion that underlies the far better known work of Kuhn (1962/1996). The incommensurability that we perceived between researchers investigating linguistic-cognitive issues and those working on the basis of sociocultural or sociocognitive views of research in L2 learning and teaching is recognized by Fleck as three differences in thought style. First, echoing Wittgenstein’s (1953/2001) concept of the language-game, Fleck recognized that the language used by members of a thought collective to describe what appears to them as reality differs from one thought collective to another. For example, the term L2 learning has different meanings according to whether researchers consider learning as the incremental accumulation of concepts that are gradually refined and combined with other concepts to form cognitive structures or, alternatively, whether they consider learning as movement along trajectories of changing engagement in discursive practices. Second, the questions that members of a thought collective ask arise from within the thought collective; in other words, the phenomena to which members of a thought collective attend are constructs of their thought style. For example, in analyzing a conversation between two people, Mackey has written in her contribution to this article that researchers working with the initial version of the interaction approach focused their attention on meaningful interaction, corrective feedback, and associated interactional features. Working within a different thought collective, although researchers may recognize the importance of interactional modifications, they may choose instead to attend to very different features, including the gender, ethnicity, and age of the participants as well as how power is exercised and resisted in interaction. The third characteristic of incommensurable thought styles, according to Fleck, is a difference in perceptions: What researchers within one thought style see, people working within a different thought style do not see, and thus they think differently about phenomena. As Fleck (1935/1979) wrote:

In L2 learning and teaching research, though researchers in different thought collectives study the same phenomenon—what Cook and Bassetti (Reference Cook and Bassetti2011) term bilingual cognition—some see it as neural activation at the interface between the cortex and the limbic system, whereas others see the genesis of cognition in cultural artifacts, social interaction, and the socioeconomic environment of individuals. Within these different thought collectives, different thought styles envisage the nature of bilingual cognition in their own and very different ways.

METHODS AND METHODOLATRY

A similar understanding of the history of ideas has been put forward by Talmy. In his contribution to this article, Talmy questions whether, in recognizing differences and even incommensurabilities between thought styles, we are not creating discursive space through a rhetoric of difference. Some of that difference may be attributed to the different methods researchers in L2 learning and teaching employ, methods that become ingrained within one thought collective but are considered esoteric and misunderstood by researchers in a different thought collective. A noticeable difference between thought styles was presented by Talmy, who recognized that, within a cognitive thought community, quantitative research is the predominant thought style, whereas researchers in a social thought community adopt a predominantly qualitative thought style. Indeed, after many years of teaching a graduate course in research methods, I have found that qualitative research methods including case study, ethnography, grounded theory, and narrative inquiry appeal to students with undergraduate majors in the humanities, whereas numerical methods involving statistical analyses of data from surveys, experiments, and language corpora appeal to students with a background in sociology or psychology. So, yes, in these student preferences I recognize a rhetoric of difference between qualitative and quantitative research methods, what Janesick (Reference Janesick2011) called different habits of mind, grounded in the personal preferences of researchers and the early training they received.

Do such differences of method lead to irreconcilable differences in understanding the phenomena of language learning and use? There are two cases showing that differences may in fact be reconcilable, and the choice of one method may lead to insights that can then be best studied by a different method. In one case, insights from qualitative research may lead to investigations best undertaken through quantitative comparisons. An example of this comes from the results of conversation analysis applied to crosslinguistic comparisons. Conversation analysts, as Schegloff (Reference Schegloff1993) argued, have eschewed quantitative analysis in the belief that when analysts count, they are not counting instances of what to the participants are necessarily the same thing. Fox et al. (Reference Fox, Wouk, Hayashi, Fincke, Sorjonen, Laakso, Hernandez and Sidnell2009), while accepting that, in same-turn self-repair, speakers are not always doing the same thing, chose to investigate the site where speakers initiate repair within the same turn. They then compared the sites at which speakers preferred to initiate same-turn self-repair across seven different languages. By means of quantitative comparisons across the seven languages, Fox and colleagues found that the sites favoring self-repair initiation within the same turn in the different languages depend to a large extent on the syllabic and morphemic properties of words in the language.

A second example of reconciling separate thought styles illustrates how a quantitative analysis may reveal broad connections between variables while, at the same time, identifying outliers whose atypical performance merits further investigation. An example of just such a quantitative study is Abrahamsson and Hyltenstam (Reference Abrahamsson and Hyltenstam2008)’s study of the relationship between language learning aptitude in adults and children and their eventual nativelikeness. Abrahamsson and Hyltenstam confirmed a correlation between quantitative measures of aptitude and nativelikeness (stronger in adults but weaker in children), but their analysis also revealed that the rare nativelike adults all turned out to be exceptionally talented language learners whose unusual ability is worth investigating further by qualitative means.

We can perhaps conclude, then, that although cognitive research is predominantly quantitative, whereas qualitative methods prevail in social research, neither quantitative nor qualitative research methods should be taken as unquestioned thought styles within their respective thought collectives. Both quantitative cognitive research and qualitative social research may produce results best investigated with a different methodology. In other words, as Talmy has argued, we should be careful that our own training and formative experiences as researchers do not lead to a habitual attachment to a single habit of mind resulting in methodolatry.

HOW DO WE KNOW WHAT WE KNOW?

If the association between thought style and thought collective is strong, then we can ask if one way to bridge the gap between social and cognitive approaches is for researchers in L2 learning and teaching in each thought collective to adopt different research methods (i.e., different thought styles). If the question of methodology is essentially “How do we know what we know?” then if we change how we know, does that alter what we know? In Lantolf’s contribution to this article, he has proposed that the knowing-how question (a question of epistemology) is not much of a question; instead he asks us first to settle what he calls the “ontological problem”—the question of what human cognition is. Strictly speaking, this is not a question of whether human cognition exists, which is an ontological problem, but rather a question of what we understand by human cognition—a problem of definition that is, in effect, closely related to the epistemological problem. In Lantolf’s terms, do we understand human cognition as occurring isolated within the brain of an individual or do we instead consider it as a dialectical relationship between the individual brain and the social world in which we live, a dialectic mediated by language, culture, and cultural artifacts? This is, then, a question of our understanding of cognition, and surely, therefore, how we arrive at that understanding is greatly influenced by the means with which we pursue it.

In the cultural-historical school of psychology that Lantolf advances, the genetic method of researching cognition (i.e., higher mental functions) is quite different from other schools of psychology. The genetic method emerged as Vygotsky’s response to the dualism of mind and body put forward by Descartes and others, and it involves studying the processes through which human cognition develops (rather than the outcome of those processes). In the genetic method, the dynamic of cognitive development can be studied in several aspects: the evolutionary development of the species (phylogenesis), the historical development of culture and tools, the development of an individual person (ontogenesis), and the continuous formative activity underlying cognition (microgenesis). Genetic investigations that aim to discover the origin, trend, rate, direction, and pattern of cognitive development describe a process very different from studying a given cognitive phenomenon or a particular behavior for its own sake. Genetic investigations are much more extensive than other investigations because they recognize that human cognition arises from interaction between biological and cultural inheritances.

To summarize, the concept of cognition espoused by Lantolf and other advocates of cognitive-historical psychology is indeed very different from the concept of cognition taken as the basis of research by psychologists who do not attend to the cultural and historical development of cognitive phenomena. The difference between the two is, I have argued, a difference of ways of knowing, of method. In other words, the difference between the two approaches is a question of epistemology, not ontology.

THOUGHT STYLES

Methods of inquiry that L2 researchers prefer differ from one thought collective to another. Because learning and using a L2 is a vast and vastly complex phenomenon, any way of investigating the phenomenon inevitably involves attending to some aspects and disattending from others. Selective attention to different phenomena is the principal way in which thought collectives differ, and it is an important part of Fleck’s theory of thought style, which he defined as

The way that different researchers attend to different aspects of L2 learning and even how the same researcher comes to attend to different aspects have been described by researchers who introspect on their own research process. In her contribution to this article, Mackey, for instance, has recounted a shift in her own attention to the nature of interaction involving language learners. From a research posture of initial disattention from the social factors underlying interaction, Mackey (e.g., in Philp and Mackey, Reference Philp, Mackey and Batstone2010) refocused her attention to consider the relationships among learners and how they impacted what learners were willing and able to listen and attend to, and thus what they produced. Mackey’s attentional change parallels the recognition by a number of authors of the different roles for the human subject in language learning experiments, depending on whether researchers focus narrowly on the phenomenon of interest or whether they expand the focus of their attention to include their subjects’ social and political roles with other subjects, with the researcher, and with the society beyond the walls of the laboratory. For instance, in her early studies of the cognitive process of reading in a foreign language, Cavalcanti (Reference Cavalcanti1983) found that her subjects not only reported their thoughts about reading but also framed them in the social context of their roles as subjects in a research study. In more recent applied linguistic studies of metaphor, which earlier studies had interpreted narrowly as an index of cognition alone, Zanotto, Cameron, and Cavalcanti (Reference Zanotto, Cameron, Cavalcanti, Zanotto, Cameron and Cavalcanti2008) argued that metaphor must be understood as social and situated, not merely a reflex of thought. Reflection by these investigators on interaction, reading, and metaphor reveals the importance of attending to the social contexts of human subjects in studies of L2 learning and teaching because they index how subjects position themselves with respect to their social and political roles with other subjects, with the researcher, and with the society beyond the walls of the classroom or laboratory. As Preston (Reference Preston1989) pointed out, even in studies of classroom interaction, researchers should attend to their subjects’ membership in different communities inside and outside the classroom.

The habits of mind—the phenomena to which researchers attend and how they analyze and interpret them—characterize a thought style. But neither thought styles nor thought collectives are immutable, and the examples of perspectival change reported by Mackey (this article), Cavalcanti (Reference Cavalcanti1983), and Zanotto et al. (Reference Zanotto, Cameron, Cavalcanti, Zanotto, Cameron and Cavalcanti2008) show how changes in the thought styles of individuals index historical development within a particular thought collective. What this implies is that, instead of looking at the differences between social approaches and cognitive approaches to research in L2 learning and teaching (the thought collectives), it is more fruitful to consider the thought styles of individual investigators, their activities, their habits of mind, and how they position themselves within the social process of investigation and publication. This is the perspective taken by Talmy in proposing an interpretivist turn in L2 inquiry, in which explaining anything says as much about the explainer as about the thing being explained.

AN INTERPRETIVIST TURN IN L2 INQUIRY

In his contribution to this article, Talmy has argued that all research in the social sciences is the outcome of the interpretive activity of particular people working from particular disciplinary and theoretical orientations in particular social contexts and historical moments. In the previous section, I have tried to address disciplinary and theoretical orientations of particular researchers in L2 learning and teaching, but let me now turn to two other issues within an interpretive perspective that have to do with the world outside the confines of the discipline: How is the work of researchers in L2 learning and teaching disseminated, and how is their work interpreted in the social world?

Nobody was born an applied linguist, and there are very few undergraduate degrees in applied linguistics, so most of us come from backgrounds in psychology, linguistics, literature, or language teaching—a diversity captured in Bob Dylan’s words, “some are mathematicians, some are carpenters’ wives” (Dylan, Reference Dylan1975, track 1). We read literature in fields that interest us and disseminate our research in journals that we read, and, as a result, the thought collective reproduces itself. In our field, one journal defines itself as “a scientific journal dedicated to the understanding of language learning broadly defined,” whereas another journal broadcasts its mission as publishing “research and discussion about the learning and teaching of foreign and second languages.” The first journal mentions science, and the second mentions teaching; and, perhaps, one journal appeals more to mathematicians, the other to carpenters’ wives. Whatever the readership, however, the economics of publishing may restrict the length of articles that any journal can print, and both journals announce on their Web sites that the length of articles they publish is generally less than 10,000 words. Such a length requirement may restrict publication vehicles for authors using interpretive methods of ethnography, case study, narrative inquiry, and even meta-analyses, while it may impose few constraints on reports of experiments, archival analyses, or other methods in which numerical representation of L2 phenomena restricts consideration of the multiple contexts in which languages are learned and limits attention to differences among learners.

A second issue raised by the interpretive approach is how research, once published and disseminated, influences the social world of those human subjects who participate in it and those who read it. As Giddens (Reference Giddens1984) first observed, a difference between the social sciences and the natural sciences is that, because the work of social scientists is embedded in its subject matter—human society—their work (intentionally or unavoidably) enters constitutively into the social phenomena they describe. In other words, whereas the natural scientist maintains a subject-object relation with phenomena, for the social scientist, the relation is a reflexive one of subject-subject. Some researchers take an active stance in the relationship with their subjects and work to uncover ways in which the context of language learning creates, reproduces, and provides opportunities for resistance to dominant power and ideology. In some cases, a researcher’s values may be fundamentally pessimistic. For example, some critical researchers identify sources of oppression and ground their analyses in the belief that we live in a fundamentally unjust world. For others, critical research is founded on the view that knowledge of the connections between local educational practices and societal ideologies may further an individual learner’s resistance and struggle against the preponderant influence of harmful public policy. Taking a less critical view but nonetheless doing research to better language teaching and learning practices is the object of action research, a form of self-reflective inquiry often undertaken by teachers to improve the effectiveness of their teaching, their understanding of their teaching practice, and the relation between their own teaching and the institutional context in which they teach.

One argument against conceiving of research in this way as an essentially pragmatic undertaking to improve the lives of language learners or the effectiveness of teachers is that such limited goals of inquiry may lead to only a partial understanding of the phenomenon. Because critical inquiry and action research are essentially normative, researchers working within these domains take both a moral and a pragmatic stance with respect to language learning. The moral stance is that certain aspects of people’s use and attitudes toward languages and their speakers are unjust, inequitable, unfair, or just plain wrong, and the pragmatic stance concerns what to do about it. Normative judgments about social situations of language use are, however, located within the political and cultural context of the critic, such that criticizing anything says as much about the critic as about the phenomenon being criticized. A second argument against research as a pragmatic undertaking is that by focusing on the solution of a specific problem, researchers may be unaware of the context of the problem and the unintended consequences of their intervention. The safest goal for critical researchers is perhaps, as suggested by Habermas (1968/1971), not to control social processes or even to influence the decisions that teachers, learners, and institutional authorities make in any determinate sort of way but rather to initiate public processes of self-reflection.

COGNITIVE AND SOCIAL LANGUAGE USAGE

Nick C. Ellis

University of Michigan

Language is essentially human. It is the crowning accomplishment of our social and cognitive competences. Language bridges society and cognition. It is a distributed, emergent phenomenon. People and language create each other, grow from each other, and act and change under the influence of each other. Language and cognition are mutually inextricable; they determine each other. Language has come to represent the world as we know it; it is grounded in our perceptual experience. Language is used to organize, process, and convey information from one person to another, from one embodied mind to another. Learning language involves determining structure from usage, and this, like learning about all other aspects of the world, involves the full scope of cognition: the remembering of utterances and episodes; the categorization of experience; the determination of patterns among and between stimuli; the generalization of conceptual schema and prototypes from exemplars; and the use of cognitive models, metaphors, analogies, and images in thinking. Language is used to focus the listener’s attention to the world; it can foreground different elements in the theatre of consciousness to potentially relate many different stories and perspectives about the same scene. What is attended is the focus of learning, and so attention controls the acquisition of language itself. The functions of language in discourse determine its usage and learning. Language structure, language acquisition, language processing and usage, and language change are similarly inseparable: They are facets of the same complex, adaptive system.

So each of us, whatever our theoretical background, shares a fascination with language, and we try to understand it. We ask our questions, so many of them, and so interesting. We do our research. Our diverse questions require different methods. Language learning diaries, functional magnetic resonance imaging (fMRI) scanners, analyses of service interactions using conversational analysis (CA), introspection, visual world eye-tracking, classroom interaction recordings, computer simulations, artificial grammar learning experiments, billion-word corpora of usage, questionnaires, dynamic assessment, event-related potentials (ERPs), think-alouds, feedback manipulations, n-back tasks, psychometric batteries, error analysis, longitudinal corpora, laboratory experiments, classroom field experiments, ethnographic research, agent-based modeling, contrastive analysis, brain connectivity analysis, dynamic systems analysis, behavioral genetics, idiographic and nomothetic approaches, thick and thin descriptions, emic and etic approaches . . . are all usefully applicable techniques, but they are useful for different things. Distrust introspection as a valid index of the working of the prefrontal cortex and anterior cingulum in bilingual code switching. There is a reason we have neuroscience. Distrust armchair analyses of language change. There is a reason we have corpus linguistics. Distrust parental recollections of their child’s language 20 years prior. There is a reason we make dense longitudinal recordings of language interaction. Distrust current computer models of emotional interaction. There is a reason we prefer people to chatbots. Distrust overzealous current technology: Do not expect rich cultural competence from computerized language instruction or solitary infants to learn language when sat in front of a TV. We are social beings. Different research methods suit different purposes. Personally, I have learned much about language from my children, from laboratory experiments, from connectionist modeling, from learning a L2, from analyzing the British National Corpus, and from behavioral genetic research, but differently so.

As we research all of our questions, we will incrementally develop a better understanding of language. But if we do it simply in the way that I have listed pieces of the enterprise here, it will be piecemeal theory. We need an additional understanding of how the pieces fit together, interacting in space and time over many different levels of granularity and timescale. Distrust any theory that claims that you can comprehensively study a component in isolation: syntax separate from lexis or semantics, form from function, representation from processing, diachronic from synchronic, knowledge from experience, behavior from brain, competence from usage, and so on (N. C. Ellis & Larsen-Freeman, Reference Ellis and Larsen-Freeman2006). Especially pertinent here is the social-cognitive gap. Such partitioning leads to theoretical, ontological, and social isolation; self-aggrandizement; and autistic hostility. Diversity is powerfully creative if there is chance of interaction (Darwin, 1859/1928; Holland, Reference Holland1998; Page, Reference Page2008, Reference Page2010). I am encouraged by the multiple perspectives currently represented within usage-based approaches to language (Behrens, Reference Behrens2009; Bybee, Reference Bybee2010; N. C. Ellis, O’Donnell, & Römer, Reference Ellis, O’Donnell and Römer2013; Robinson & Ellis, Reference Robinson and Ellis2008; Tomasello, Reference Tomasello2003; Trousdale & Hoffmann, Reference Trousdale and Hoffmann2013). These hold that we learn language while engaging in communication—the “interpersonal communicative and cognitive processes that everywhere and always shape language” (Slobin, Reference Slobin and Slobin1997, p. 267). Some of the basic tenets, many of them explicitly addressed by de Saussure (1916/1983), include the following:

1. 1. Language is intrinsically symbolic, constituted by a structured inventory of constructions as conventionalized form-meaning pairings used for communicative purposes.

2. 2. Language is intrinsically linked to human cognition and processes of perception, attention, learning, categorization, schematization, and memory.

3. 3. Adult language knowledge consists of a continuum of linguistic constructions of different levels of complexity and abstraction. Constructions can comprise concrete and particular items (as in words and idioms), more abstract classes of items (as in word classes and abstract constructions), or complex combinations of concrete and abstract pieces of language (as mixed constructions). No rigid separation exists between lexis and grammar.

4. 4. Constructions may be simultaneously represented and stored in multiple forms, at various levels of abstraction (e.g., concrete item: table + s = tables and [Noun] + [morpheme -s] = plural things).

5. 5. Constructions can thus be meaningful linguistic symbols in their own right, existing independently of particular lexical items. Nevertheless, constructions and the particular lexical tokens that occupy them attract each other, and grammar and lexis are inseparable.

6. 6. Language structure emerges ontogenetically from usage in particular contexts. Development is slow and gradual, moving from an initial reliance on concrete items to more abstract linguistic schemata. This process is dependent on the type and token frequencies with which particular constructions appear in the input. Storage of wholes depends on token frequency; the development of abstract linguistic schema depends on type frequency.

From analyses of large usage corpora, we can analyze the latent structures of language and their roles in the associative and cognitive learning of language (N. C. Ellis et al., Reference Ellis, O’Donnell and Römer2013). This is the stuff of cognitive psychology, associative learning theory, and corpus linguistics. But in addition to construction forms and their frequencies, there are their meanings—embodied, attended, consciously formed in dialogue and dialectic, and situated and encultured in social and educational interaction. Usage-based theories hold that an individual’s creative linguistic competence emerges from the collaboration of the memories of all the meaningful interactions in his or her entire history of language usage.

Cognitive linguistics (Croft & Cruise, Reference Croft and Cruise2004; Langacker, Reference Langacker, Barlow and Kemmer2000; Robinson & Ellis, Reference Robinson and Ellis2008) provides detailed analyses of how language is grounded in our experience and our physical embodiment, which represents the world in particular ways. The meaning of words in languages and how they can be used in combination depends on the perception and categorization of the real world around us. Because we constantly observe and play an active role in this world, we know a great deal about the entities of which it consists. This familiarity is reflected in the nature of language. Ultimately, everything we know is organized and related to our other knowledge in some meaningful way, and everything we perceive is affected by our perceptual apparatus and our high-level motor control and motor apparatus: Spatial language is grounded in the visual processing system as it relates to motor action, the multiple constraints relating to object knowledge, dynamic-kinematic routines, and functional geometric analyses (Coventry & Garrod, Reference Coventry and Garrod2004). Action elements relating to hand posture, joint motions, force, aspect, and goals are all relevant to linguistic distinctions (Bergen & Chang, Reference Bergen, Chang, Trousdale and Hoffmann2012). Meanings are embodied and dynamic (Spivey, Reference Spivey2006); they are flexibly constructed online. Here we have the embodied, dynamic mind of modern cognitive science (A. Clark, Reference Clark1998; Rosch, Varela, & Thompson, Reference Rosch, Varela and Thompson1991).

Shared attention, shared cooperative activity, and shared cognition are key to meaningful language usage. In their first 2 years, infants develop their capabilities of attention detection (gaze following), attention manipulation (directive pointing), intention understanding (the realization that others are goal directed), and social coordination with shared intentionality (engaging in joint activities with shared interest, negotiating meanings), and these processes are central in child language acquisition (Tomasello, Reference Tomasello1999, Reference Tomasello2008).

The nature of language follows from its role in social interaction. Social interactions are typically characterized by what philosophers of action call shared cooperative activity (Bratman, Reference Bratman1992) or joint actions (H. H. Clark, Reference Clark1996). Joint actions are dependent on what might be broadly called shared cognition, a human being’s recognition that she can share beliefs and intentions with other humans. Thus usage-based approaches emphasize how language is learned from the participatory experience of processing language during embodied interaction in social and cultural contexts in which individually desired outcomes are goals to be achieved by communicating intentions, concepts, and meaning with others. Conversation partners scaffold and coconstruct meanings. Socially scaffolded noticing (Schmidt, Reference Schmidt1990) solves Quine’s (Reference Quine1969) problem of referential indeterminacy and builds so much more. The dynamics of language learning are inextricably linked to the dynamics of consciousness, in neural activity and in the social world as well (U. Frith & Frith, Reference Frith and Frith2010). Input to the associative network is gated by consciousness, and consciousness is coconstructed in social interaction (N. C. Ellis, Reference Ellis2005; C. Frith, Reference Frith2010). In these ways, the input to the associative network is socially gated (Kuhl, Reference Kuhl2007).

Language—and language learning—is ever thus. It takes place in social usage, involving action, reaction, collaborative interaction, intersubjectivity, and mutually assisted performance (Lantolf & Thorne, Reference Lantolf and Thorne2006). Speech, speakers, identity, and social relationships are inseparable (Lee, Mikesell, Joaquin, Mates, & Schumann, Reference Lee, Mikesell, Joaquin, Mates and Schumann2009; Norton, Reference Norton1997; Tarone, Reference Tarone2007). Activity theory emphasizes how individual learning is an emergent, holistic property of a dynamic system comprising many influences: social, individual, and contextual. Action provides a context within which the individual and society, mental functioning, and sociocultural context can be understood as interrelated moments (Wertsch, Reference Wertsch1998). The associative learning network is culturally gated. Tomasello’s constructionist approach to language unites with his research in comparative primate cognition, the unique place of social cooperation in humans, and the Vygotskian intelligence hypothesis, whereby regular participation in cooperative, cultural interactions during ontogeny leads children to construct uniquely powerful forms of perspectival cognitive representation, including language itself (Moll & Tomasello, Reference Moll and Tomasello2007). The last 40 years have seen huge progress in research into social cognition, and within social cognitive neuroscience there is now a rich understanding of the role of implicit and explicit knowledge in social cognition (C. D. Frith & Frith, Reference Frith and Frith2008), of the role of consciousness and metacognition in social interaction (C. Frith, Reference Frith2010), and of the brain mechanisms involved in these processes (U. Frith & Frith, Reference Frith and Frith2010). From the very name of the field, you can tell that there is no social-cognitive divide within contemporary social cognitive neuroscience.

How should we research this usage? If language learning is in the social cognitive linguistic moment, we need to capture all these moments so that we can objectively study them. We need large, dense, longitudinal corpora of language use with audio, video, transcriptions, and multiple layers of annotation for data sharing in open archives. We need these in a sufficiently dense mass that we can chart learners’ usage history and their development (Tomasello & Stahl, Reference Tomasello and Stahl2004). We need them in sufficient detail that we can get down to the fine detail of CA analyses of the moment (Kasper & Wagner, Reference Kasper, Wagner and Atkinson2011). Brian MacWhinney has long been working toward these ends, first with the Child Language Data Exchange System (CHILDES; MacWhinney, Reference MacWhinney1991) and then with Talkbank (MacWhinney, Reference MacWhinney2007). These projects have developed various computerized language analysis (CLAN) tools for computer analyses of large bodies of data, right down to, in collaboration with Johannes Wagner, tools for a fine-grained CA bank (http://talkbank.org/CABank). With these types of data, we can study the cognitive and the social. This way the future lies. These are huge contributions to language acquisition research, though, at the moment, the data are relatively sparse. We need much more, especially for studies of L2 learning and teaching. We need big, dense, longitudinal data sets that we can all observe; it is up to the field to contribute to these open archives.

We need theoretical integration, too. Cognition; consciousness; experience; embodiment; brain; self; and human interaction, society, culture, and history are all inextricably intertwined in rich, complex, and dynamic ways in language. We require additional perspectives on dynamic interactions between levels, perspectives provided by approaches such as complex adaptive systems (N. C. Ellis & Larsen-Freeman, Reference Ellis and Larsen-Freeman2009), dynamic systems theory (de Bot, Lowie, & Verspoor, Reference de Bot, Lowie and Verspoor2007; N. C. Ellis, Reference Ellis2008; Spivey, Reference Spivey2006), and emergentism (N. C. Ellis, Reference Ellis1998; N. C. Ellis & Larsen-Freeman, Reference Ellis and Larsen-Freeman2006; MacWhinney, Reference MacWhinney1999).

At the AAAL 2013 colloquium in Dallas, I was asked to talk from a cognitive perspective along with Martha Bigelow, who was presenting for the social side. I knew how lucky I was in this pairing: Language Learning had previously published her excellent monograph Mogadishu on the Mississippi: Language, Racialized Identity, and Education in a New Land (Bigelow, Reference Bigelow2010). At the end of her presentation, I was asked if I wanted to respond. Postperformance zonked and worried by the adequacy of my own contribution and the large, varied, and potentially awesomely critical audience (you, maybe?), I was not prepared for this. But our joint reaction was easy. In the moment, we faced each other, checked the trust in each other’s eyes, smiled, shook hands, and embraced. It emerged in that cognitive-social moment: There was no cognitive-social divide. For my part, it was easy to embrace the position of someone who blends together approaches from focus groups, sociocultural analysis, storytelling, psycholinguistics, and educational evaluation, as they are appropriate in dealing with very real educational and social issues of immigration. As Ortega noted in her discussion of our presentations at the colloquium, we can go beyond simple dichotomies if we choose to imagine better, we can strategically try to undo these outdated notions, and we can socially coconstruct something superior. There is no social-cognitive gap.

BLENDING SOCIAL AND COGNITIVE RESEARCH TRADITIONS IN LANGUAGE LEARNING AND TEACHING

THE COGNITIVE-SOCIAL GAP

RESPONDING POSITIVELY TO THE GAP: BRIDGE BUILDING AND BRIDGE TRAVELING

For Ellis, language and language learning and use are at once cognitive and social. Therefore, their study must be tackled from both cognitive and social perspectives. In other words, it is the very nature of the objects of study that creates the justification for bridging cognitive and social inquiry; the need for bridge building is thus rationalized as an ontological imperative. Ellis stipulates, however, a disciplinary division of labor: Multiple social and cognitive dimensions of language learning and use dictate different granularities and timescales for research and thus require different research questions and demand different methods. This being so, cognitive and social dimensions must be addressed by different communities—or in Fleck’s (1935/1979) sense, by different thought collectives and their resulting thought styles. Thus there is epistemic independence in the complementarity Ellis proposes. He also sees a need for theoretical integration and finds it in usage-based approaches. This family of theories stipulates that the cognitive and the social, the abstract and the physical, and even the empirical and the meditative create layers of phenomena that, at the metatheoretical level, are interdependent and must be integrated into explanations of human cognition (MacWhinney, Reference MacWhinney1999; Varela, Thompson, & Rosch, Reference Varela, Thompson and Rosch1991). When complementarity in epistemic independence is looked at from this integrative metatheoretical perspective, a bridge is built that invites the traveler to cease viewing the cognitive-social gap as a threat.

For Bigelow, too, the relevant phenomena to be explained include social and cognitive dimensions. She feels, however, accountable to bridging the social and the cognitive by blending the two dimensions in her own research praxis—that is, within single studies and within the same research program. Thus, her account bears witness to researchers who respond to the gap by trying to close it through bridge traveling. Bigelow shares her arresting awareness of the vulnerabilities and risks involved in these efforts. For senior researchers like her, and even more so for junior researchers and students, bridge traveling via epistemic blending, she tells readers, can be a tremendous challenge. One could argue that the motivation for epistemic blending of such proportions goes beyond ontological understandings of the object of study and is fueled by a bottom-up, problem-oriented enticement (or felt necessity) to transcend, in practice, whatever cognitive-social dichotomies appear to be in opposition. In Bigelow’s case, studying the underserved population of Somali adolescents in urban Minnesota challenged what she and her colleagues thought they knew about language learning and use (Bigelow & Tarone, Reference Bigelow and Tarone2004) and what they decided they needed to know, and it commanded a broaching of social and cognitive dimensions of the problem at once. Her bottom-up justification for the need to travel cognitive-social gaps resonates with views of applied linguists who argue that research should begin with an analysis of both the given problem that needs to be understood and transformed and the anticipated consequences of the given research, with the choice of methods and theories subordinated to that analysis. Far from being monolithic, these views are epistemologically diverse and range from critical pragmatism to critical realism, from critical pedagogy to critical theory, and even to poststructuralism (see discussions in Bygate, Reference Bygate2005; Corson, Reference Corson1997; Crookes, Reference Crookes2005; McNamara, Reference McNamara2012; Pennycook, Reference Pennycook2010).

NOT TWO BUT MULTIPLE UNDERSTANDINGS OF GAP BRIDGING

Ontological bridges might be built to look quite different. Ellis’s is one example. Lantolf, in his contribution to this article, offers a quite different example, choosing a different theory that helps build another ontological bridge: Vygotskian sociocultural theory. His choice of theory helps bridge the gap organically and naturally, to the point that no gap is felt, and no bridge is needed. This is done not through complementarity, however, but by recognizing that the dialectic between social activity and mental activity is mediated by language and cultural artifacts. The aspiration to undo ontological dualisms and dichotomies is shared by Vygotskian sociocultural psychology and by usage-based approaches (see N. C. Ellis & Larsen-Freeman, Reference Ellis and Larsen-Freeman2006). The dualisms, however, are overcome by different means, either at the intratheoretical or at the metatheoretical level.

Bridge traveling, or the integration of the cognitive and the social within a single study or within the same research program, can also occur in different ways. Bigelow’s account is one example. Mackey, in her contribution to this article, offers a different case when she recounts how her research practices and overall research program changed to better balance social and cognitive factors. In Mackey’s case, the changes seem to have been prompted by theoretical pressures brought about by the social turn in the field of SLA (Block, Reference Block2003). Mackey’s process is perhaps best described by Fleck’s (1935/1979) notion of stylized remodeling (discussed by Young in this article). No researcher or research community can forever be insulated from surrounding intellectual developments. Eventually, he or she will be changed by varying degrees of familiarity with those developments. Theoretical pressures to bridge the gap such as those underlying Mackey’s case are very different from the bottom-up, real-world pressures described by Bigelow.

QUESTIONS OF (IN)COMMENSURABILITY, ALWAYS PENDING

I have created two metaphors, bridge building and bridge traveling, to characterize the positions taken by Ellis and Bigelow in this article. This rhetorical move allows me to suggest that it is possible for individual researchers to learn from felt cognitive-social tensions and to act on them positively across different epistemic lines along the shores of the gap. I have also suggested that there are not one or two but several ways to respond positively to the gap, as in the parallels and differences between Ellis and Lantolf in their contributions, on the one hand, and between Bigelow and Mackey in theirs, on the other. Yet, with these two rhetorical moves, I would not want to minimize important questions and tensions that run through the contributions in this article under the theme of (in)commensurability. Just how possible are mutual understanding and successful communication across different research communities that stand at different sides of the cognitive-social gap? In thinking about this question, it is important to remember that views regarding (in)commensurability are greatly diverse across relevant fields, such as the philosophy of science; history of ideas; or science, technology, and social studies. I choose to discuss here three positions toward the problem of (in)commensurability that have been distinguished by Holbrook (Reference Holbrook2013).

A well-known model of interdisciplinary communication is offered by Habermas (1981/1984), who argued that commensurability, or at least reciprocal understanding, is a tall order, though possible through the human capacity to engage in rational, action-oriented deliberation. The best-known position overall, however, is Kuhn (1962/1996), who was inclined toward incommensurability. Communication attempts that build on the shaky grounds of translation (always less satisfying than access to the original), argued Kuhn, will result in nonunderstanding or, worse, in overestimation of understanding and thus misunderstanding. The only solution would be for researchers from different communities to learn one another’s discourses, though developing such interdiscursive competence may prove extremely difficult, if not impossible.

As an alternative to the Habermasian and Kuhnian renditions of the dilemma of (in)commensurability, Holbrook (Reference Holbrook2013) offers a third option, the reflective invention model, crafted from his reading of Bataille (1957/1993) and Lyotard (1983/1988). This model predicts that attempts at establishing understanding across epistemic boundaries will eventually lead to communicative breakdown precisely because each interlocutor, by virtue of being a member of a different thought collective, holds different assumptions and plays by different rules of the game of knowledge construction. (This is the perspective emphasized in the contributions to this article by DeKeyser, Hulstijn, Talmy, and Young.) The breakdown point marks the recognition of belonging to different epistemic discourses. And it is at the point of recognition of difference that the response becomes a critical choice:

I find the Bataille-Lyotard reflective invention model poignant and useful. Through this lens, (in)commensurability can be seen as a local, emergent achievement whose centripetal impulse is located in critical moments afforded by a recognition of difference, which are then resolved either by reaffirming one’s own epistemological identity and intellectual affiliations and thereby retreating from the possibility of communication (i.e., choosing incommensurability) or by transcending those differences in some way and keeping up the hard work of achieving at least some degree of success in communication (i.e., choosing commensurability).

Let us assume, then, that moments of recognition of difference and imminent communication breakdown are opportunities either to lock into incommensurability or to choose to work toward commensurability. What might be conditions that can sway individual researchers to choose to work toward commensurability? I would like to put forth two tentative answers. One draws from the rationalist traditions that predominate in the philosophy of science and that are likely to be familiar to many readers. The other looks for inspiration into poststructuralist traditions in postcolonial studies and feminist studies that may be less familiar to many readers. I include both as epistemological translations of each other, fully cognizant of their distinct potential to reach and persuade different readers.

RATIONALIST AND POSTSTRUCTURALIST INROADS INTO COMMENSURABILITY

From the rationalist tradition of philosophy of science, it is relevant to search for intellectual predispositions that may support commensurability. Understanding seems a necessary but not sufficient condition to achieve commensurability. Once the difference is recognized and understood, one can always opt out of communication efforts. Nor is knowledge an antidote for incommensurability because the deeper one’s knowledge of someone else’s thought collective becomes, the more acutely one recognizes the limits of superficial translation across thought styles. This may, in itself, lead to hopelessness and communicative disengagement at the edges of the gap. Respect for epistemological and methodological difference is also likely to be insufficient. For example, there is respect in Ellis’s bridge-building proposal when he exhorts researchers to interact across disciplines and to abhor intellectual isolationism. But the question remains: Can researchers truly communicate when they are members of complementary yet independent research communities?

Tolerance of ambiguity, trust, and self-reflexivity may be among the needed ingredients, according to a panel of natural and social scientists who were charged by the European Science Foundation to evaluate felicitous conditions for interdisciplinary collaborations (ESF Standing Committee for the Social Sciences, 2013). Tolerance of ambiguity is required in communication with colleagues “whose research might, for example, look dangerously reductive or trivially broad from the point of view of their own epistemic community” (ESF Standing Committee for the Social Sciences, 2013, p. 21). In turn, such ambiguities are likely more tolerable, the panel concluded, when trust among individual researchers is strong. Being able and willing to trust is often associated with self-reflexivity, as “researchers frequently seek out trustworthy others, and also become willing to trust, only after critical reflection on the qualities of, and gaps within, their own disciplines and methods” (ESF Standing Committee for the Social Sciences, 2013, p. 21). This element of self-reflexivity is important in the interpretive turn called for by Talmy in his contribution to this article.

Support for commensurability can also be found in poststructuralist thinking. From this perspective, feelings of inner tension and threats to researcher identity can be uniquely illuminated by a willingness to confront questions of moral values. As social scientists like Kirschner (Reference Kirschner2013) put it, this requires going well beyond cold rationalism and drawing from nontraditional, visceral explorations of ideologies and ethical consequences of knowledge. The key condition supporting commensurability from the poststructuralist perspective is a willingness to transcend dichotomous epistemic boundaries. Two good candidates for doing this are the construct of an interstitial view offered by postcolonial literary theorist Bhabha (1994/2004) and the notion of “world”-traveling proposed by feminist philosopher Lugones (Reference Lugones2003).

Bhabha (1994/2004) proposed that human beings have the capacity to achieve an interstitial view when, in the unfolding of an experience of marginality, they manage to displace the dichotomous assumptions implicit in a difference they recognize and suffer; by doing so they are able to reach beyond the other-imposed and self-imposed dichotomy and assert their right to difference in equality. Bhabha explained: “It is in the emergence of the interstices—the overlap and displacement of domains of difference—that the intersubjective and collective experiences of nationness, community interest, or cultural value are negotiated” (p. 2). Lugones (Reference Lugones2003) proposed that human beings have the capacity to travel symbolically to someone else’s world when we “understand what it is to be them and what it is to be ourselves in their eyes” (p. 97). This “world”-traveling in turn facilitates empathy and the recognition of the other as a valuable, legitimate interlocutor because “only when we have traveled to each other’s ‘worlds’ are we fully subjects to each other” (Lugones, Reference Lugones2003, p. 97); by cultivating “world”-traveling, resistance to oppression and coalitions across different kinds of marginality can be achieved. In other words, complex communication is achieved (Lugones, Reference Lugones2006).

HOPEFUL COMMENSURABILITIES

Ultimately, the two answers to the challenge of cultivating commensurability I have offered are epistemological translations of each other. Readers will have to judge how superficial or successful the translation work is. As in all translations, my purpose in considering both has been one of reaching across linguistic and epistemological boundaries.

Rationalist answers to the conundrum of (in)commensurability, such as those offered by the ESF Standing Committee for the Social Sciences (2013), are helpful. Understanding, knowledge, respect, tolerance of ambiguity, trust, and self-reflexivity may all be important resources that individual researchers should cultivate if they are interested in closing the cognitive-social gap in the study of language learning and use. But these rationalist explanations are also somewhat limited: These virtues do not explain what ignites some researchers’ desire to achieve mutual understanding and communication, particularly when the ideals of bridge building and bridge traveling may risk disciplinary identities, as Holbrook (Reference Holbrook2013) puts it and as Bigelow suggests on the basis of her own experience. How, when, and why, for example, would one even know that the trouble is worthwhile? More visceral analyses such as those offered by Bhabha (1994/2004) or Lugones (Reference Lugones2003) offer helpful ways of deepening our appraisal of (in)commensurability as an emergent achievement of the Bataille-Lyotard kind (Holbrook, Reference Holbrook2013). It is through liminal, performative processes such as the interstitial view and “world”-traveling that the recognition of difference during communication across epistemic communities may end up being seized as a momentary opportunity for working at commensurability. In critical rationalist epistemologies, these emotional-affective, relational-empathic, and moral-political accounts of commensurability fall outside the proper domain of knowledge and science. However, in a wide range of constructivist, structuralist, and poststructuralist epistemologies, knowledge is viewed as value laden and ideological. There is, thus, more willingness for researchers to feel they are accountable to axiological dimensions of knowledge and to acknowledge that they are influenced by them in their research praxis.

In the end, however, it should be clear that commensurability cannot be thought of as a constant virtue, only as an emergent achievement. Complete, permanent commensurability is not possible, nor is it desirable. For one, ways of thinking and knowing are not monolithic but encompass considerable internal heterogeneity, both across sides of any perceived gap and within the same thought collective (Bronson & Watson-Gegeo, Reference Bronson, Watson-Gegeo, Duff and Hornberger2008; Philips, Reference Philips2004). Moreover, the preservation and coexistence of a degree of incommensurability within the field, or the cultivation of epistemological pluralism, is a good thing (Ortega, Reference Ortega and Atkinson2011). It may be wiser, therefore, to think of commensurabilities, in plural, as strategically chosen acts whose value is contingent on local purposes and dynamic conditions, acts that are always unfinished and ongoing, in a precarious but hopeful temporary balance.

EDITORS’ CLOSING THOUGHTS

Looking back at the contributions to the original 2013 AAAL colloquium, the lively and positive audience interaction they inspired, and the resulting formulations gathered in this article, we believe we can legitimately draw the following two conclusions. First, contributors agree that there is no single, monolithic social-cognitive gap in L2 learning and teaching research, or at least there need not or ought not to be a gap. The contributions gathered in this article succeed in disentangling various independent contrasts in ontology, epistemology, and methodology, distinguishing a range of views rather than a single dichotomy. Furthermore, whereas it is true that, as one member of the colloquium audience remarked, when a gap is bridged there is still a gap, we note that genuine attempts to build bridges constitute attempts to understand views on the other side, which cannot but lead one to reflect on one’s own stance. We trust that the contributions to the colloquium and to this article have shown many areas for dialogue and collaboration.