Criticisms
The methodology of discourse analysis has been subject to some criticisms specifically related to the methods, assumptions, and limitations of the methodology itself. While these are often inherent, there are often ways to mitigate their liabilities. What follows is a summary of some of the criticisms of this methodology, as well as possible solutions, or at least ways to lessen or acknowledge their effects.
Small Data Sets
Data sets used for discourse analysis can be and often are smaller than expected of many other types of research, especially those looking for statistically-supported external validity. Instead, a deeper, rather than broader, analysis of the key terms and their uses is coded not only by number of occurrences, but also by relevance. Thus, more focus is put on the key terms themselves, including their nuances, implications, and assumptions than simply the number of utterances. This attention further limits the scope of enquiry, as only texts that are rich sources of data are used, rather than full texts. Though this criticism that data sets are too limited has some basis, it would seem this choice is more one of focus than omission, as experiments in the psychology of perception, a closely related field, are valid with as few as one subject. The reason for this validity, as in discourse analysis, is that the perceptions of subjects, regardless of how many have them, are real (Tonkiss, 2012, p. 376).
Limited External Validity
Due in part to the acceptance of small data sets as outlined above, the external validity of discourse analysis findings has been challenged. Though statistically unsupportive of external validity and application, theoretical generalisation can be attributed to findings. This is possible and reasonable as discourse analysis relies heavily on logical interpretation and assertion, rather than observed fact. Though the limits of theoretical findings are obvious, they do provide a starting point for not only discussion, but action. Moreover, usually specific social settings and phenomena are studied with this method, so the need for widely generalizable external validity may not be necessary.
“Examining Processes” Which Are Fluid
An emergent nature is inherent not only to this type of research, but also to the perceptions upon which it is based. The changing nature of perceptions, conceptions, and language not only contribute to limiting external validity, but also to the internal validity of the research at hand. The processes which create meaning are particularly vulnerable to time, which may make evolving definitions difficult to categorize. Therefore, the goal of this type of research is not often to look for specific solutions, but rather to understand the problem and processes that may lead to possible solutions (Tonkiss, 2012, p. 378).
Researcher’s Language Preconceptions Working with or against Subjects'
Challenges with the fluidity of language definitions are compounded by the fact that they still must be communicated, and therefore interpreted by at least two people. This is a serious criticism, as, when situated in a postmodern understanding of constructed meaning, the analysis itself may double the variables being analysed. Even at a more basic level, clusters of meaning may appear in the data, but may not have the correct or intended meaning attributed. The strongest offset to this effect may also be exactly where the difficulty lies: in the logical analysis of the text. If totally reliant on the conjecture of the researcher, this has the possibility of veering sharply from the intended message of the subject. If, however, safeguards such as using subject-defined terminology, cross-referencing within and between texts, explicit and agreed definitions, and strong and clear reflexivity are used, this danger can be effectively reduced.
“Expert Languages”
The challenge of language perceptions and variation can be exacerbated by the complexity and specificity of ‘expert languages’, or jargon specific to a field or function. This type of language is rife with implications and assumptions, beyond the challenges of simple understanding if the researcher is not ‘in-group’. However, this is rarely the case, as researchers often have a personal interest, and some induction, into the subject being studied. In this case, subject-specific terminology can actually be a benefit, as it can be a form of shorthand, with clearly and explicitly defined definitions that have been cross-referenced and reviewed within and by the group at large (Tonkiss, 2012, p. 375).
Small Data Sets
Data sets used for discourse analysis can be and often are smaller than expected of many other types of research, especially those looking for statistically-supported external validity. Instead, a deeper, rather than broader, analysis of the key terms and their uses is coded not only by number of occurrences, but also by relevance. Thus, more focus is put on the key terms themselves, including their nuances, implications, and assumptions than simply the number of utterances. This attention further limits the scope of enquiry, as only texts that are rich sources of data are used, rather than full texts. Though this criticism that data sets are too limited has some basis, it would seem this choice is more one of focus than omission, as experiments in the psychology of perception, a closely related field, are valid with as few as one subject. The reason for this validity, as in discourse analysis, is that the perceptions of subjects, regardless of how many have them, are real (Tonkiss, 2012, p. 376).
Limited External Validity
Due in part to the acceptance of small data sets as outlined above, the external validity of discourse analysis findings has been challenged. Though statistically unsupportive of external validity and application, theoretical generalisation can be attributed to findings. This is possible and reasonable as discourse analysis relies heavily on logical interpretation and assertion, rather than observed fact. Though the limits of theoretical findings are obvious, they do provide a starting point for not only discussion, but action. Moreover, usually specific social settings and phenomena are studied with this method, so the need for widely generalizable external validity may not be necessary.
“Examining Processes” Which Are Fluid
An emergent nature is inherent not only to this type of research, but also to the perceptions upon which it is based. The changing nature of perceptions, conceptions, and language not only contribute to limiting external validity, but also to the internal validity of the research at hand. The processes which create meaning are particularly vulnerable to time, which may make evolving definitions difficult to categorize. Therefore, the goal of this type of research is not often to look for specific solutions, but rather to understand the problem and processes that may lead to possible solutions (Tonkiss, 2012, p. 378).
Researcher’s Language Preconceptions Working with or against Subjects'
Challenges with the fluidity of language definitions are compounded by the fact that they still must be communicated, and therefore interpreted by at least two people. This is a serious criticism, as, when situated in a postmodern understanding of constructed meaning, the analysis itself may double the variables being analysed. Even at a more basic level, clusters of meaning may appear in the data, but may not have the correct or intended meaning attributed. The strongest offset to this effect may also be exactly where the difficulty lies: in the logical analysis of the text. If totally reliant on the conjecture of the researcher, this has the possibility of veering sharply from the intended message of the subject. If, however, safeguards such as using subject-defined terminology, cross-referencing within and between texts, explicit and agreed definitions, and strong and clear reflexivity are used, this danger can be effectively reduced.
“Expert Languages”
The challenge of language perceptions and variation can be exacerbated by the complexity and specificity of ‘expert languages’, or jargon specific to a field or function. This type of language is rife with implications and assumptions, beyond the challenges of simple understanding if the researcher is not ‘in-group’. However, this is rarely the case, as researchers often have a personal interest, and some induction, into the subject being studied. In this case, subject-specific terminology can actually be a benefit, as it can be a form of shorthand, with clearly and explicitly defined definitions that have been cross-referenced and reviewed within and by the group at large (Tonkiss, 2012, p. 375).