Conclusion
The ideal research method for the investigation of speech acts, and in particular for the investigation of compliments, does not exist. There is not even a method that is in a general way better than all the others. In this paper, I have argued that an assessment of a particular method always depends on the specific research question that the researcher tries to answer because the different methods vary enormously in their suitability for specific research questions. One particular method may provide reasonably reliable results for one specific question or set of questions while it is of little value for another question or another set of questions.
Thus, I fundamentally disagree with Manes and Wolfson’s assurance about their diary method, which they call “ethnographic approach” (Patterns – Field – Diary):
It is our conviction that an ethnographic approach is the only reliable method for collecting data about the way compliments, or indeed, any other speech act function in everyday interactions. Manes and Wolfson (1981: 115).
I have categorized the different research methods on the basis of the distinction proposed by Clark and Bangerter (2004), who proposed the terms “armchair”, “field” and “laboratory” for three types of research methods on the basis of the prototypical locations in which they are carried out. It is typical for many linguists to propose one of these methods as the only reliable method for linguistic research and to exclude all the others. In this paper, I have taken the view that all the three methods and the subtypes that I have introduced have their intrinsic values but all of them are limited in the types of research questions that they can tackle.
I am not proposing that every researcher who wants to investigate compliments has to employ all the research methods proposed above in order to get a comprehensive picture. Instead I want to argue that researchers should adopt a more modest attitude in their discussion of the chosen research method. The methods of their choice may be the best possible for the very specific research question that they are asking but other methods are equally valid for different questions.
The methods also differ in the generalisations that they allow. It is an essential part of academic argumentation to generalise results and to claim a wider application of ones findings than just the specific examples or sets of examples that have been analysed. However, such generalisations often need to be more modest, too. Fictional data, for instance, is usually shunned by linguists because the findings based on fictional data cannot be generalised to everyday natural conversations. It is true that such a generalisation is not possible. But if linguists and pragmaticists turn their interest from their narrow focus on everyday natural conversation to language use in general, all instances of language use have to be seen as situated. The totality of language use is made up of many, many different varieties and subvarieties of spoken and written language. They all have to be seen within their special conditions and limitations, and research results based on one variety should not be generalised too easily to other varieties.