Researching language learning: theories, evidence, claims
Epistemological assumptions and ILL research
'Untying the bundle': the 'variables' approach in ILL research
These insights have implications for research in the variable analysis tradition, so at this point we examine claims which are made for 'the variable' as an important component in traditional ILL research. Hatch and Farhady (1982), for example, explain that '[Height, sex, nationality, and language group membership are all variables commonly assigned to people' (p.12), going On to distinguish between attributes like these, which are 'of the all-or-nothing type', because 'the Ss [subjects] either are or they aren't foreign students; they are or they aren't speakers of French .. .', and other variables which are susceptible to gradation and quantification: 'In other cases we can measure very accurately how much of the variable subjects possess' (ibid.). (See similarly Hatch and Lazaraton 1991.) Although this approach is by no means ubiquitous in ILL research, significant numbers of studies continue to be published which appear to be heavily influenced by a variable analysis perspective, and as an illustration of this broad orientation we have selected the specific research questions from just three recently published examples.
1. What are the relationships between a range of language learner variables (e.g. previous language proficiency, motivation, beliefs about and strategies for language learning) and the language achievement of advanced Chinese EFL learners at tertiary level; and
2. What differences relating to these variables exist between relatively successful and unsuccessful learners? (Wen and Johnson 1997: 28)
This study is principally an attempt to provide more detailed basic descriptive data on language proficiency, language behavior, and language attitudes as a function of the immigration background for a small sample ... of high-school students in a rural community in Northern California. (Hakuta and D'Andrea 1992: 73)
The objective of this investigation was to determine the influence of supplementary mother-tongue instruction, self-esteem, length of residence, and schools, on the French second language performance of first-generation immigrant children. (Fazio and Stevens 1994: 433--4)
In each case, a language learning 'problem' is identified, and a number of factors are considered as potentially having a bearing on that problem. Investigations take the form of measurements of these factors and identification of their respective significance. The expression of each problem incorporates a fairly precise delineation of who is involved ('advanced ChineseEFL learners at tertiary level', 'a small sample ... of high-school students in a rural community in Northern California', 'first generation immigrant children'), and what the language learning enterprise is ('French second language performance', '[English] language achievement'). The specificity is consistent with the prescription outlined by Cook (1986), who demonstrates how an 'overall, untestable proposition about speech and writing' such as 'Is speech more important than writing?' is progressively 'narrow(ed] down ... so that it can receive a reliable answer'. The 'precise, answerable question' he devises is: 'Are the scores on the EPVT of a group of educated French adult learners of English in technical schools in France who are taught orally significantly better than those of an otherwise identical group who are taught through writing?' (p.19). The gap between these two questions offers us a preliminary indication of one of the reasons why, in our view, variables-based ILL research has so far not delivered on its promise of progress towards consistent, aggregated findings. Byrne (2002: 20) puts the realist position well when he points out that:
... the implications of localism/context as expressed in interaction, mean that no single hypothesis can be generalized beyond the exact conditions under which it is tested. If there is any non-linearity then no covering law is ever possible. In this frame of reference experiments merely describe local and unique conjunctions. Hypotheses cannot be somehow summed into a general overarching account.
