A social realist approach to research in applied linguistics
The limits of the empirical in applied linguistic research
Just as we have noted that the majority of the questions we have cited as representing applied linguistic research are likely to be most obviously located in the domain of situated activity, we also suggest that the kind of evidence which is most obviously adduced in answering such questions is empirical evidence. Thus the researcher who wants to know how speakers of a minority language feel about its preservation may interview some of them as a source of data; a comparison of different teaching methods will probably collect test results from students in Group A and those in Group B; the researcher who is interested in coherence in the 'long turns' which candidates are required to produce as part of their test of English language proficiency is likely to collect examples of this speech in the form of recordings and transcripts.
Such data sets are empirical in the sense that what they contain is observable, but of course the real interests of the researchers are frequently in phenomena which are unobservable, for which the data have to serve as traces. So it is important to remember that data which are thought to be 'empirical' are often something other than a collection of instances of the phenomenon under investigation.
Interviews about people's beliefs and attitudes are not only problematic for researchers in being themselves both a research method and a form of social interaction; they also must be deemed to convert something which can never be directly perceived (an attitude) into something which can (a statement or response). This is one of the strengths of ethnography as a means of finding out what people actually do (when and how do they use the variety?) as opposed to what they say they believe (do you think Variety X should be on the school curriculum?). Even so, the behavior of the speakers under observation still provides evidence only of the empirical where the distinction drawn in realist accounts between the empirical, the actual and the real. The implications of this include the fact that accounts which reduce a phenomenon to its empirical manifestations will be seriously mis leading. This can be illustrated by reference to the language learners whose test scores are analyzed by the researcher comparing two or more teaching methods. By our account, these scores can only be traces of language proficiency: they are not that proficiency in themselves.
Measurement and theory are thus imbricated in a powerful way. For the positivist, measurement provides direct, unmediated information about the world, and its problems lie largely in the area of technical accuracy. From our position, by contrast, measurement must always be understood as a theoretical tool, one by means of which we 'translate' the world into scientifically significant descriptions. It is on the basis of these that we then develop causal accounts. A recognition that '... measurement in social research will always be ... an act of translation' (Pawson 1989: 287) entails a rejection of the notion that measurement is a direct description of the social world - or indeed of the natural world either. One example used by Pawson is the measurement of temperature. The temperature of a physical substance cannot be directly observed, but changes in its temperature are measured indirectly, by observing changes in a measuring instrument, such as a thermometer, which have been devised by previous scientists to 'translate' the presence of heat into something else which is visible, such as the movement of mercury along a marked tube. In response to the recognition that 'measurement is not simply a matter of observation but also of conceptualization' (ibid.: 40), there are those who advocate abandoning all forms of empirical measurement as necessarily partial, relativistic and 'socially constructed'. Our position lies somewhere between these two extremes, besieged, as Byrne (2002: 15) provocatively expresses it, by two opposing 'gangs', 'the reductionists' and 'the innumerates':
The first gang can count but don't know what they are counting, why they are counting, or what to do with what they have counted when they have counted it. The second can't count, won't count, and assert that counting is a vile and perverse activity which ought not to be allowed.
We seek approaches to research which take account of the empirical, but are not limited to it. From this position, we point out a further limitation of empirical data, namely that there are many phenomena of interest to the applied linguist which are inherently refractory to empirical investigation. These include those aspects of the social, cultural, linguistic world which we have described as 'relational' above. To take another example, and one which has a bearing on many branches of social research, how can empirical data shed light on questions about the relationship between social class membership and differential educational attainment?
The initial problem which confronts the researcher who is interested in the significance of social class within some social process, such as benefiting from an education program, for example, is how 'class' is to be rendered in practical research terms. It is not directly observable - a person's social class membership is not something the researcher can see - so how does he or she utilize class when studying phenomena such as, for instance, making progress towards learning an additional language? Two of the strategies most common across the social sciences draw so heavily on the empirical that they are arguably empiricist, though in different ways.
The better-established approach is to theorize class as an attribute, something which people 'have'. Quantitative studies in this tradition may either seek to 'control for' social class, by including only subjects whose measure of this attribute is deemed to be within the same kind of range, or they may investigate how the possession of different degrees of the attribute correlates with different outcomes. This approach can be illustrated by reference to one example of an applied linguistic study in the quantitative tradition. In the research by Fazio and Stevens (1994), social class was identified as potentially relevant in differential patterns of achievement in French by 'minority children' immigrating to Canada. The schools attended by the children were classified as being of 'low' or 'middle-high' socioeconomic status. This variable 'was determined by the Four Factor Index of Social Status developed by Hollingshead (1975). The four factors used in the calculation of SES are: occupation ...; education ...; sex ...; and marital status ..." (p.426). Our disagreement with this kind of approach is that it takes variate traces of complex social relations and conflates these with the relations themselves. There is then a loss of ontological depth as the distinction between the world as it is and the world as it appears to us is collapsed. This is not to argue that a study which sets out to measure the correlations between levels of income, say, and indications of educational attainment has no value. It is to claim, though, that the approach taken in such studies does not constitute a measure of social class: it may provide us with traces of the operation of class processes, but the significant challenge in research terms is to connect the traces with the processes, and this means moving beyond the empirical.
The alternative approach we might classify as 'experiential'. 'Class', in this younger tradition, may be theorized not so much as a given attribute as a possession which people use to perform in the world. An example from applied linguistics is provided by Rampton (2002) in his analysis of young people deploying differently valorized accentual varieties of English, in this case stylizations of 'Cockney' versus 'posh', to convey both 'class awareness' and 'class critique'. Again, insofar as this approach reduces social class to what speakers do and are aware of - which, we might add, provides us with a different sort of 'trace' - it has the same kind of effect as the first approach discussed. But now it is experience which is taken as the empirical element — the means by which we know class.
Thus, while the quantitative approach indicates traces of structural distributions in the patterns it identifies, it does not provide evidence of how social class relations are experienced by people themselves. By contrast, the interpretive approach indicates how social actors experience social class processes, but has little to say about the structural features of these. What the approaches share, however, is an emphasis on empirical data, be they aggregated in the form of levels of income or individualized in the form of utterances spoken.
As we have indicated, applied linguistics shares with many other social science disciplines a need to respond to wider debates about how we can know the social world, and, in broad terms, it also participates in the struggle between methodologies which are more inclined towards measurement and those which are more inclined towards interpretation. Debates in the journals suggest that there can be a productive relationship between theory and practice, including both theory about applied linguistics and applied linguistic practices (such as language teaching), and theories about the nature of the social world and the practical challenge of how it may be researched. A recent review of health-care communication as 'a problematic site for applied linguistic research', for example (Candlin and Candlin 2003), reports on how qualitative and quantitative approaches have been combined, and how applied linguists concerned with the study of health-care communication have drawn on studies undertaken by researchers in other academic disciplines as well as those by health-care practitioners in the course of their own work.
Our own position stresses the links between theory and the empirical world, seeking to steer between an approach in which the world tells us, as it were, what theories to have, so that 'the whole thrust and meaning of theory is irresistibly driven by the empirical world as it appears to our senses' (Layder 1993: 61) and an alternative approach which suggests that the connection between the empirical world and theory formation is arbitrary or contingent; that is that theories can be of little help with developing knowledge of the empirical world. Our belief that the world extends beyond our empirical knowledge of it does not imply or entail a reduction of the empirical simply to a portal through which access to 'deeper reality' is enabled. The empirical, in other words, is no less an element of reality for being directly present in ways that other elements of the social world are not. Our central point here is that forms of measurement and classification are critical to research which seeks to explicate the connections between the empirical, the actual and the real dimensions of social relations. The aim, then, is an elaboration of the kind of '... theory which is guided rather than limited by empirical evidence'
