A social realist approach to research in applied linguistics
The centrality of theory in applied linguistic research
Many of the sets of data (interview and questionnaire responses, test scores, recordings of speech in assessment interviews and so on) are not in themselves the phenomena we are trying to investigate, and yet these data sets are useful for research. This is because the empirical data to which we do have access are suffused with theories about categories, measurement and knowledge. All too often, however, these theories are not made explicit, and the emphasis in research design is on the technical adequacy of the questionnaire, sample size or consistency in interview style among researchers, rather than on the theoretical adequacy of the research categories themselves.
Let us explore this a little further with reference to research into differential achievement among language learners. Since learners' motivation is often taken to be a relevant variable in such studies, we will consider briefly the ways in which 'motivation' has been represented in applied linguistic research to illustrate the issue of theoretical concepts. The characteristics of motivation have been discussed, as 'a learner trait' or with reference to 'situated language identity theory' (Noels and Clement 1996 in Dornyei 2001); researchers debate how accurately different sub-components of motivation have been identified, how to distinguish between goals, motives and orientations, and so on. To exemplify this area, we cite one study into ESL acquisition in Quebec. Belmechri and Hummel (1998) generated a 'Motivational Intensity Score' for each student in their study, and used:
(a) Factor analyses to determine which orientations clustered together to form general orientations in this particular context, and (b) after delineating general orientations, a multiple regression analysis to find out which combination of these best predicted motivation, (p.227)
They conclude that 'orientations explained 41 per cent of the variance in motivation' (p.239). The study has many of the characteristics of the traditional approach to SLA research, including the identification of variables, the quantification of their relative weighting, and a quest for some generalizable findings about the nature of motivation in SLA. This goal is evident in much of the relevant literature, so that Crookes and Schmidt, for example, 'suggest .., that a theory of the role of motivation in SL learning ought to be general and not restricted to particular contexts or groups' (1991: 502). On the other hand, the researchers in our example study counsel caution about their findings, 'because our sample was not fully representative ... and was not large enough to be conclusive' (Belmechri and Hummel 1998: 241). Moreover, motivation researchers increasingly recognize the importance of 'context':
We would certainly not dispute that language learning takes place within a social context, nor that socially grounded attitudes may provide important support (or lack thereof) for motivation. We do not claim that there are no interesting relationships among social contexts, individual attitudes, and motivation.
However, this position (perhaps revealingly constructed with several double negatives) does not seem to us to resolve the tension inherent within a desire to specify salient variables and draw generalizable conclusions about them, while simultaneously acknowledging the specificity of time, place and social location of the language learners focused on in the research. These authors declare that a desirable goal for 'a theory of the role of motivation in SL learning' is that it be 'general', but the underlying, more implicit, theoretical proposition is that the label 'motivation' names something in the world which can be directly investigated. We suggest that similar problems surround studies which take many superficially unproblematic concepts for granted. These include ethnic categories, but also concepts such as * discourse coherence', 'Japanese/ Chinese/English requesting/apology styles', and, in a generic way, the studies involving comparisons between 'learners in Group A' and 'learners in Group B', where membership of the group may mean being taught in a specific country, or by a specific method, or any other supposed 'variable' in the pedagogic experience.
The realist social theorist Byrne (2002: 31) has dramatically declared 'death to the variable - or rather', as he continues:
let us understand clearly, once and for all, that variables don't exist. They are not real. What exists are complex systems, which systems are nested, intersecting, which involve both the social and the natural, and which are subject to modification on the basis of human action, both individual and social.
In other words, the crucial search is for ways of accessing - both theoretically and empirically - the relational nature of social (including linguistic) phenomena, which is emphatically not the same quest as that for the elusive, definitive list of interacting factors which might, if properly specified and isolated, account for the behavior, experience and texts in which applied linguistic researchers are interested.
