Researching language learning: theories, evidence, claims
Social realism and empirical research
It is probably fair to say that the longest-established and still dominant tradition in ILL research (Lazaraton 1995) involves the following kind of approach, as summarized by Lightbown (2000: 438) in her review of SLA research between 1985 and 2000:
The specific goals of the various research projects differ, but there is a unifying desire to identify and better understand the roles of the different participants in classroom interaction, the impact that certain types of instruction may have on FLiSL learning, and the factors which promote or inhibit learning.
Much of what constitutes this kind of research tends to be consistent with conventional, successionist models of causality. Successionist models usually take the form of identifying a dependent and an independent variable and then proposing a hypothesis which suggests that they will vary inversely or conversely, thus enabling some form of causal inference to be drawn. Ellis (1990: 199) summarizes this approach thus:
The L2 classroom researcher seeks to show how instructional events cause or impede the acquisition of a second language. In order to achieve this, it is necessary to (1) identify which instructional events are significant, (2) find valid and reliable measures of the L2 learning that takes place and (3) be able to demonstrate that the relationship between instructional events and learning is in some ways causal.
Although critiques of the successionist concept of causality have been made in the ILL literature (e.g. Block 1996; van Lier 1990), and we will discuss alternatives later, it remains a dominant assumption in a significant proportion of the literature.
ILL research, then, like most research in both natural and social science, explores why things happen. Our own position in response to this question is a 'modified materialist naturalist' one. Therefore, our starting point in understanding why things happen is a belief that the world consists of phenomena- including human beings and social structures- which have distinctive properties and powers. These objects exist independently of our understandings of them. (This points to a significant limitation in the purely interpretivist research which is often seen as the alternative to successionist accounts of causality, as we explain below.) It is the combination of the powers and properties of human beings and social structures which generates the experienced empirical world. The attribution of these properties and powers is not arbitrary, but is given in what phenomena are: it is given in the nature of human beings, for example, that they are mortal, that they have the biological attributes which enable them to use language, and so on; social structures are emergent, durable, and have the facility to frustrate or further the projects of people both individually and collectively.
This implies a radically different view of causality and therefore of research from that found in both successionist and interpretivist traditions. From a realist point of view, rather than merely cataloguing repetitions of regularly occurring co-events, science should seek to understand and identify the causal relations or mechanisms which produce the observed, empirical regularities. Sayer (2001) summarizes the key components of a realist view of causality:
Causes- that is whatever produces change- should be understood as causal powers possessed by objects (including individuals and social structures) that mayor may not be activated. Whether they are depends on contingently related conditions, and if and when they are activated, what results also depends on contingently related conditions. (p.968)
So what are the things in the social world which have these properties and powers? The components of the social world, from a realist perspective, are human beings and the products of their social interaction, including social structures and culture. Itis the interaction between these components of the social world, and the realization of their properties and powers in particular settings and combinations, that give rise to the problems with which social research- including applied linguistic research- is concerned. This account of 'why things happen' allows for the occurrence of patterns and regularities, without entailing a commitment to a nomothetic perspective of 'governing laws'. Regularities- such as, say, tendencies for groups of students with certain characteristics to outperform groups of students with different characteristics on particular tests of L2 performance may be indicative of causal relations, and establishing such patterns therefore often entails undertaking quantitative research. The account also allows for a recognition of the powers of human agency, including the reflexivity which can act as a 'confounding variable' in traditional process-product studies. The position is also marked by seeking to distinguish between necessity and contingency in causal relations. Research of this type will not discover 'universal laws', because the actualization of the properties and powers of the objects in the world is context-dependent, but it will aim to generalize beyond the individual case of the single ethnographic study, by drawing on particular conceptualizations of propensity and probability.
We can now turn to a richer image of 'complexity' than that suggested when the term is used, as it sometimes is, as a synonym for 'complicated'. It is common to make reference to the 'complexity' of the process of using different languages in different social contexts, and of learning how to do so. For example, Lightbown and Spada (2001: 42) identify a number of 'variables' which 'have been found to influence second language learning'. These include 'intelligence, aptitude, personality and motivational characteristics, learner preferences, and age'. Some of the difficulties for researchers, they suggest, arise because 'these learner characteristics are not independent of one another: learner variables interact in complex ways. So far, researchers know very little about the nature of these complex interactions.' Similarly, Cook (1986: 13) characterizes 'the real world' as 'a complex bundle of many things', but concludes that the researcher's task is therefore to- as it were- untie this bundle and extract separate 'things' to measure.
However, recognizing that variables interact is not the same as recognizing that in their interaction they generate emergent, irreducible phenomena, which are themselves capable of interacting back on their constituent elements. In this view of complexity, there is an emphasis on the fact that 'the behavior of complex systems arises from the interaction of its components or agents' (Larsen-Freeman 1997: 143, emphasis added). Larsen-Freeman is one of the few researchers who have begun to explore more fully the implications of rejecting successionist accounts of causation, in characterizing SLA as a complex nonlinear process. If this is what it is, she argues, 'we will never be able to identify, let alone measure, all of the factors accurately. And even if we could, we would still be unable to predict the outcome of their combination' (ibid.: 157). As she points out, researchers in the domain of neuroscience are now modelling the brain as an example of a complex nonlinear system, and developments in this area contribute to explanations of extremely sophisticated mental processes. Features of the brain which qualify it to be thought of in this way include its 'decentralized' character, and its feedback systems which contribute to growth and self-regulation (Johnson 2001). The characterization of language itself as a cultural emergent property entails a recognition of its complex and emergent features (Beaugrande 1997, 1999). And language acquisition is also recognized as an emergent phenomenon by researchers who concentrate on the internal cognitive dimension of the process. As Ellis (1998) puts it:
Emergentists believe that simple learning mechanisms, operating in and across the human systems for perception, motor-action and cognition as they are exposed to language data as part of a communicatively-rich human social environment by an organism eager to exploit the functionality of language, suffice to drive the emergence of complex language representations. (p.B57)
Ellis and Larsen-Freeman both advocate studying language and language learning in relation to other areas of enquiry, and both point to the need for researchers to pool resources across disciplines, since complex problems are unlikely to be solved by simple techniques. However, we are so far not aware of very much applied linguistic research which has taken On these developments to the point of recognizing emergence and complexity across the full strata of the 'domains' of the social world, although we shall discuss later the work of some who are contributing to such developments.
