This paper uses a discourse analytic technique of reflexivity to revisit my own unease around the use of thematic content analysis (hereafter TCA) in my recently completed Masters dissertation. This dissertation -pieces of which have been published and/or discussed elsewhere - used Foucauldian theory and discourse analysis (hereafter DA) to examine the operations and effects of advice columns. The TCA study was intentioned as an initial descriptive sweep of the content of sampled advice columns, but the siting of this extremely labour-intensive work in the dissertation became increasingly fraught with paradigm and/or philosophical dilemmas. Eventually, rather ignominiously, the TCA study was savagely edited and self-consciously squeezed into a chapter on "methodology", whereafter many pages were devoted to discussion and deconstruction of its presence! The results of the TCA study will be set out briefly for the sake of concreteness, but the focus of this paper is methodological: to explore what TCA is/does; to allude to the tensions between TCA and DA; and to evaluate TCA's efficacy as a "more democratic" qualitative method which avoids, for example, the slippery philosophical / theoretical roots of DA, and the jargon it requires and produces.
What is TCA?
Danziger offered several informal guidelines on how to do TCA. It involves a fairly flexible, semantic unit of analysis (e.g. a word, phrase, sentence, paragraph, etc.) based on decisions about themes of "meaning". This differs from more traditional (i.e. statistical) approaches to content analysis which rely on fixed or stable units of analysis, e.g. word counts, etc. It is understood that each protocol (e.g. a text as a demarcated piece of discourse) could contain many themes. However, within TCA, each semantic unit may "fit" more than one theme which obviates the rigorous either/or categorizations normally associated with content analytic methods. On the surface then, this is not antithetical to DA's valorization of indexicality, variation and contradiction, or indeed, to Potter & Wetherell's (1987) definition of DA as 'sifting out recurrent themes of meaning'.
TCA and advice columns
Eighteen content-based themes were developed inductively through close examination of the realms of experience / meaning represented as "problematic" in the sample of published advice texts. These themes were also informed (but not prescribed) by categories used in previous content analytic studies of advice columns (e.g. McFadyean, 1988; Weinberg, 1989; Kurtz, 1990). Themes were not equally sized, constant units: some were broader than others and they were not internally homogenous, i.e. several "strands" might be discernible within each. It is noted that the word "theme" was intentionally used to convey a common-sense, descriptive sweep of content, and to distinguish it conceptually from "discourses" in the Foucauldian sense, i.e. productive, transformative, normalizing, institutionalized effects of ideology and power/knowledge (cf. Parker, 1992).
Brief descriptions of the themes are provided below. Themes 2 to 10 refer to "relationship problems".
Reliability of thematic coding - percentage of inter-coder agreement - was calculated on a randomly selected 10% of the sampled texts by an independent coder. Overall, agreement was high, 91.84%. In positivist discourse, this statistic means that an "acceptable" degree of consistency of categorization, using the above themes, was achieved. Discourse analytic perspectives ironicize such consistency as indicative of any kind of "truth", and I will return to this point in a later evaluative section.
Briefly, some of the quantificatory findings are described below. First, the advice column was confirmed as a "feminine genre" in that 81% of letter writers were female, 10% male and 9% of unspecified gender. Feminist media critics have argued that, ideologically, this reproduces women as needful of expert assistance with being women. However, this finding should be understood in the context of the sampling in this study, i.e. Fair Lady and Femina are predominantly (white) women's magazines, and You magazine a "general interest" publication (50:50 male to female readership). Men's magazines might include differentially focused sites of advice (e.g. computer information!), and reflect higher proportions of male letter-authorship to themes forums.
Second, the findings from content analytic studies in British women's magazines, that young women formed the predominant group of advice-seekers (even in magazines targeted at older women), was confirmed in this study. The modal range for sampled advice seekers was 15-19 years (mean was 17.83 years, with youngest letter-writer being 9 years and the oldest, 79 years). This age group is thought to be confronted, within the white, middle class readership supposedly represented, with various developmental issues relating to, for example, sex and relationships, or career-planning and study, while still being financially dependent on parents. Thus, advice columns provide confidential, professional, free advice which would be unavailable through other formal channels.
Third, evaluation of my tallied frequencies of themes 2 to 10 (i.e. "relationship problems") against other content analytic studies on advice columns support the dominance in the women's advice genre of representations of "romantic love" and "relationship" (e.g. emotional / sexual / love problems). What is interesting, however, is that the thematic distributions show complexities of specialization of content in the different sites of advice in the sampled magazines. Please refer to graphic distributions of frequencies herewith: TABLES 2, 3 & 4.
At least two trends become visible in this distribution format. The first is the emergence of the "generally focused" advice column, e.g. Ask Elizabeth Duncan in Fair Lady magazine, Share your Problems with Susan James in Femina, and Janet Harding's Lifeline in You magazine. This is evidenced by a fairly equable spread of all 18 themes within these forums, i.e. they field general, ordinary problems with everyday living / loving.
The second trend is evident in graphs of Fair Lady and Femina magazines, which run 3 and 4 concurrent sites of advice respectively. Here, women's experience is carved up into problematic fields, usually associated with an "expert" of some persuasion to solve crises with particular knowledges and techniques. These columns are characterised by almost monothematic content, e.g. Beauty Clinic in Femina (almost exclusive focus on problems around "Physical attractiveness"), Dr David Delvin's Sexpert's Guide in Femina (which focuses on sex and medicalized information) and Couple Clinic in Femina (which focuses on a variety of "relationship crises". This suggests, of course, the intervention of fairly rigorous editing and selection (or even, horror of horrors, hoax letters or the writing of letters by editors themselves!), which pose major threats to the validity of TCA findings within a positivist framework. And it is to these issues of evaluation that I now turn.
The usefulnesses of TCA
First, TCA's flexibility permits it to be used to amplify other kinds of analysis. Thus, my usage (above) is pragmatic or functionalist in the sense that my intention was never to use TCA as an "end" or "final word" by itself. It provides a static, exhaustive, descriptive sweep of issues represented in advice texts, which "orders" discourse to facilitate more focused analysis of the discourses at work in particular areas and in selected texts, e.g. "monogamy" in different sites of advice. The prevalence / frequencies of represented issues - cloaked as they are in positivist discourse - can be taken forward in arguments about hegemony and/or operations and effects of power/knowledge as institutions which "muscle in" on particular realms of problematic experience, e.g. medicalization of sexual knowledge or psychologization of relationship-maintenance. Furthermore, this might inform further levels of intervention and/or political appropriation, e.g. taking up particular problematic issues, or the marginalization thereof, with magazine editors, publishing critical writings on the normalizing operations of information-dissemination, etc.
Second, TCA studies, perhaps because they sustain their positivist roots, are able to produce comparable and evaluative data to that in existing mainstream literature on advice columns. This provides a more globalizing view of the advice genre. In a merely descriptive or uncritical vein, this results in the piling up of endless content analytic studies which have attested to the preponderance of "love problems" in advice columns; or in the documented historical changes in content of problems from "etiquette issues" in the 1940's and 1950's (e.g. "who pays for the flowers at my wedding?"), to the explicit sexual information requested in recent times. The latter is taken as empirical evidence of permissiveness, sexual freedom and enlightenment rather than as proliferating forms of institutional surveillance and disciplinary power, reproduced through a technology of confession (Foucault, 1977). Once again, the TCA study might provide material which may be taken forward in ways that serve ideology-critique and empowerment, at social and subjective levels.
Third, TCA assumes an "atheoretical" - or "theory-free" - stance, i.e. it is not burdened with slippery philosophical / theoretical underpinnings and overtones or jargon. It sets out an easy recipe to follow or adapt, and since method and content are fairly distinct, its applications to any form of written / spoken discourse, are limitless. This democratization of method reproduces TCA as extremely user-friendly, and enters the realm of "common sense", i.e. categorization into themes is something we all do every day of our lives. This combines well with participatory / feminist research approaches where the data / participants / texts are allowed to "speak", giving voice to particular issues. Furthermore, themes may be developed collectively among participants in group discussions, or fed back to communities for further discussion and commentary.
Within this vein, and quite curiously perhaps, the support of TCA by the powerful, so-called scientific, positivist edifice, bolsters the "truth" and "reality", and "incontestability" often, of findings produced. Thus, there is a sense that what is really going on is captured - even more so if themes are collectively developed or if reliability coefficients are high. This was particularly disturbing in the ways in which my TCA study was lauded, and became a form of self-congratulation, by editorial staff at Fair Lady and Femina. The accessibility and mainstream acceptance of TCA served as an empirically viable means of "excluding" the ruder noises of critical discourse stories.
You don't have to be a discourse analyst to see the problems with TCA!
In the above section, I have explored some ways in which TCA might be reproduced as an anti-dote or panacea for the oft-cited elitism, subjectivity and opacity (= inaccessibility) of DA. In this concluding section, I reflexively explore some reasons for why my TCA of advice columns made me uneasy. The main problem is, of course, that TCA and DA straddle oppositional approaches to and assumptions about (a) language, and (b) the relationship between methodology and truth (i.e. goals of research). They are, thus, incompatible at a number of levels; they occupy different camps. Rather than a militarist "shooting down" of the one by the other, however, DA might be used, through its interrogatory destabilization / rupture / subversion tactics, to "strengthen" the kinds of TCA currently produced.
The first problem which becomes evident with TCA from a (jaundiced) discourse analyst's view is the atheoretical position adopted. This produces "findings" which say at the levels of "description", without explanatory or critical edge. TCA treats discourse as neutral or transparent, i.e. we can look through it to examine "content" or "meaning" underneath. This is presented as an unproblematically logical process of 'making sense' of discourse(s), of extracting information. From within a Foucauldian approach, Parker (1995) warns of the difficulties with such "logical" processes of reducing or reifying particular kinds of "sense" or information from large amounts of 'unmanageable' discourse. The context and complexity of material may disappear or be hidden.
Advice columns are a case to point where rigorous editing ensures seamless, unidimensional content. TCA describes manifest content, which cannot address the ways in which language constructs, transforms or reproduces experience through power/knowledge operations in institutions. How do power, knowledge, ideology, institutions, discourses impinge on and shape content / meaning? How are the realms of experience (e.g. love and sex, emotion and financial problems, monogamy and marriage counselling, etc.) implicated and interwoven within one another in productive ways?
The second problem which becomes evident within a DA appraisal of TCA is that the methodological flexibility of TCA relegates it to a murky space between quantitative and qualitative research. This antagonism between hegemonic positivism and resistant ethnomethodology pits rigour - as the standard - against the complete anarchy of subjectivity, and worse, intellectual flabbiness. That TCA panders to positivist discourse (e.g. representivity of samples, reliability, validity, objectivity, issues of boas, etc.) means that it claims to produce "accurate" and "consistent" estimates of "truth". TCA has no tradition which produces a critique of the positivist model of science; therefore, it cannot reflect critically on the kinds of knowledges it produces.
Again, my TCA study of advice columns is a case to point where inter-coder reliability coefficient of 91.84% was achieved. TCA does not provide tools to interrogate this statistic. For example, another researcher might produce different themes of equable strength and persuasiveness; or the consistency of these themes does not reflect socio-political realities of the white middle classes as claimed, but rather reflects editorial policy on selection of issues deemed to be representative of the target group.
Lastly, considering the above problems, I believe that it is very difficult to do "good" TCA research. Given the lack of theory, it has been seen as a "soft option" which attracts those who either can't do, or spurn through ignorance, statistical rigour. At worst, this group incorporates those individuals who have neither theoretical / methodological map, nor political agenda. This produces research which is atheoretical, descriptive (and therefore reproductive) of a status quo, and/or irrelevant to the social context in which it occurs. What is needed within TCA is sharp, critical argument and what Burman (1991) has termed "political appropriation", i.e. a moral-political and activist agenda. Findings need to be linked with what has preceded them, and with the social context which has produced them. While DA can use the lesson from TCA on democratization of methodology, TCA could be strengthened by a critical comparison with DA.