Methodology in Cognitive Linguistics
Why Cognitive Linguistics is not yet as empirical as might be
Reasons of practice
Two practical factors may be invoked to explain why Cognitive Linguistics has not yet progressed in the direction of empirical research to the full extent that might perhaps be expected on the basis of its theoretical position: the fact that training in experimental techniques or corpus research is not a standard part of curricula in linguistics, and the recognition that the “public image” that Cognitive Linguistics presents to young researchers and newcomers does not contain a commitment to empirical research as a conspicuous feature.
The first factor is undoubtedly the major one: coming from traditional curricula in language and linguistics, many cognitive linguists are simply at a loss on how to get started on empirical research of an experimental or a corpus-based kind. (And, to repeat, the present paper will not immediately help them, because it was conceived as a metatheoretical overview and not as a practical guide.) The practical steps to be taken are indeed numerous and often cumbersome. In the case of corpus research, you need to collect appropriate corpus materials, retrieve the relevant observations from the corpus, code the observations for all the features that you think might be relevant in the analysis, perform the right quantitative analysis and then start your second round in the empirical cycle with an interpretative effort. In the case of experimental research, the practical difficulties are similar: you need to set up an experimental design with the right stimuli (something which you may achieve only after an initial pilot study), find an appropriate set of experimental subjects, carry out the actual experiments, analyze them with the relevant techniques– and then start your second round in the empirical cycle with an interpretative effort.
It is no surprise then, that linguists trained in a more purely analytic paradigm need time and help to get acquainted with this form of doing research. At the same time, there are clear indications that many within the Cognitive Linguistics community are willing and waiting to receive such an initiation. A good example is the appeal of the “collostructional” approach introduced by Stefanowitsch and Gries (2003). It applies existing methods of statistical collocation analysis (i.e. methods that measure the strength of association that exists between two elements) to the relationship between the lexical items that occupy the slots in a construction and the construction as such. Cast in a readily applicable format, the method has soon attracted a number of scholars working within Construction Grammar: see Gries and Stefanowitsch (2006) for a sample. From the perspective here, the appeal of the collostructional method shows that there is indeed a demand for the adoption of empirical techniques– thus strengthening the assumption that a lack of familiarity is one of the main explanations for the situation.
At the same time, the collostructional example reveals a potential danger: not much would be achieved if cognitive linguists, out of ignorance of alternatives, would simply equate empirical research with readily available specific techniques like the collostructional analysis. What we should achieve, in the long run, is for cognitive linguists to be aware of the full range of empirical techniques, experimental as well as corpus-based. But that will require a reorganization of the average curriculum and a long-term investment in empirical training.
A second practical factor that should be mentioned involves the example provided by the leading authorities in Cognitive Linguistics: what is the image that newcomers and young researchers get when they are confronted with the work of the most highly valued researchers in the field? What methodological impression does a novice get from Cognitive Linguistics if he or she looks at the current work of the founding fathers? Their work, in fact, is to a large extent of an analytical, theory-building kind that does not go, in terms of methodological procedure, beyond the traditional methodology of contemporary linguistics: remember that Table 2 in Room for expansion in the use of an empirical methodology, showed a more conservative picture for the plenary talks at ICLC9 than for the general session talks. Let’s assume that we restrict the set of founding fathers of Cognitive Linguistics to Langacker, Lakoff, Talmy, and Fillmore. (The position of Fillmore may need some explanation, because sociologically speaking, he never identified very strongly with Cognitive Linguistics as a movement. His ideas about frame semantics and construction grammar constitute an integral part of the foundational concepts of Cognitive Linguistics, however.)
Now, in their work on grammatical phenomena and linguistic theory, Langacker (in spite of the fact that he is the person who coined the phrase “usage-based model”: Langacker 1988), Talmy, and Fauconnier are highly traditional linguists in terms of data gathering and data analysis: no systematic corpus analysis or ingenious experimentation in this line of research. Fillmore is a notable exception, to the extent that the FrameNet project that pursues a large-scale application of frame semantics to the English lexicon is definitely and deliberately corpus-based.
Lakoff too occupies a specific position, to the extent that the “cognitive commitment” that he formulated in Lakoff (1990) has led him to numerous forms of interdisciplinary collaborations with philosophers, literary scholars, and other cognitive scientists. In particular, he lies at the basis of a neural theory of language that looks at the neurophysiological basis of his Conceptual Metaphor Theory. However, the way in which Conceptual Metaphor Theory is usually practiced is largely an introspective matter (see Haser 2005, and Gibbs, this volume, for a critical assessment). It is, in fact, possibly the easiest form of linguistics that you can do in the domain of Cognitive Linguistics: collect a number of expressions for an arbitrary target field in a given language, identify a common semantic denominator for them, and conclude that thàt is the way the target field is conceptualized in the language, in other words, that thàt is the way people think in that language.
The picture is perhaps slightly caricatural, but the restrictions on this type of work will be clear. To really establish the conclusion, one would first have to assess the importance of the expression in the language: what other expressions are there, how often are they used, and what factors actually determine the choice between the alternatives? Such questions require a thorough investigation of actual usage, and in any case a methodology that goes beyond just enumerating examples. Next, concluding from the mere presence of a form in the language to a certain way of thinking requires at least that the productivity of the image inherent in the message is established, i.e. that it is made clear that the metaphor is not a dead one but actually influences or reflects, to a greater or a lesser extent, thinking. To be explicit, the point made here is not that attempts to establish an effect of language on thinking by means of adequate techniques do not exist, but rather that such an approach is far from dominant in research in the realm of Conceptual Metaphor Theory.
In other words, what is probably the most broadly attractive and most widely known form of research in Cognitive Linguistics is at the same time one of the most facile ones in methodological terms, in spite of the long-standing efforts of people like Ray Gibbs (and more recently, researchers like Lera Boroditsky, 2001) to provide empirical underpinnings for Conceptual Metaphor Theory. To the extent that this contributes to a certain image of what it involves to do interesting work in Cognitive Linguistics, this is not advantageous for the expansion of empirical methods in Cognitive Linguistics.
