Limitations of the approach

I have applied in this research a discursive and rhetorical approach. Th e rhetorical approach highlights the argumentative context as well as the way in which some arguments are made to look plausible, while others

are implicitly or explicitly contrasted and rejected. As a consequence, the fi ndings of the analyses and the discussion of results are rather divided in tone. My intention has not been to argue that the experience of old age and health is dichotomous, or nothing but argumentation between two opposite poles. Instead, the occasionally dualist structure has served the purpose of clarifying the ambiguous nature of the old age experience. My approach has not been strictly either theory-driven or data-driven. My interest in diff erent discourses arose from readings of the literature and from a closer analysis of the biographical interviews, before I had decided on the topic for my fi rst sub-study. What initially inspired my interest in the biographical interviews was the way that the interviewees listed their daily activities when they were asked about their age or health. Th ese interests continued throughout the research process, which drew the focus of analysis towards these topics.

Having said that, my approach was certainly data-driven in the sense that I studied in close detail how diff erent views were put forward, and I have tried to follow Silverman’s rule (1993) to test my own analyses and fi ndings by paying attention to deviant cases. Also, the collection of another dataset through group discussions provided an interesting alternative data source.

If interviews are conducive to creating a question-response pattern and in this way to reinforcing the dualist image, the group discussions produced more complex and nuanced data that shed interesting light on how diff erent views were negotiated and contrasted. Indeed, from this perspective, group discussion data off er an important contrast to face-to-face interviews.

Th e biographical data were collected in 1995–96 and the group discussions a few years ago. One may well ask whether a dataset collected today would reveal very diff erent ways of talking about old age and health?

My position on this is clear: the topics I have addressed in this research, i.e.

the meaning of old age, the ambiguity of old age and health identity, the morality of health talk, healthism and the question of older people’s agency and position in society have defi nitely not lost any of their importance;

quite the contrary. Indeed, as is indicated by some recent studies, activity has become an increasingly prominent keyword in ageing research. Likewise, the discussion and debate about individual responsibility for health and healthist ideas has only gathered momentum. In short, I very much suspect that the views put forward in my analyses would be even more prominent today. With the growing number of pensioners in our society, older people’s diff erent roles as grandparents and as mentors in working life and life after retirement has received even more attention than before. In other words I would venture to guess that the identities and positions assigned to old age and available for older people have continued to diversify.

Th e aim in qualitative research is not to produce statistical generalizations, but rather to consider whether and how the results are transferable and

applicable to another context (Silverman 2005). One way of doing this is to weigh and compare one’s results against earlier and current research and against results from other societies and cultures, which is precisely what I have tried to do in this research. Th is is one way of enhancing the validity and reliability of the research and at the same time of giving the reader the chance to assess these. Comparing one’s results with previous research also serves the purpose of improving validity and reliability. Furthermore, it is important to show that the data have been studied critically and systematically and that contrasting and “deviant” cases have been taken into account and used to test one’s own interpretations (Silverman 2005). In the case of empirical qualitative research this requires that the data collected and the analysis are shown to the readers so that they can judge for themselves the plausibility and rigour of the analysis (ibid.). Rather than using the data simply to dress the analysis, for instance by inserting brief exemplary quotations, it is better to use larger data segments and to illustrate in detail how the analysis has been conducted. Th is is not easy when the research is published in article format where editorial rules set specifi c limits on the size of data extracts and on the detail of analysis. In this study, however, I have included large data extracts and shown in detail how the analysis was conducted. I have also included both more common and less common ways of talk about particular topics.

Th e use of transcribed and translated data also raises questions about the actual data source. Is it the original spoken interview, or the transcribed text, or the translation data that ends up in the research articles? According to Nikander (2008b), the validity of research is enhanced by the use of audio or videotaped data that allow for multiple hearings, and showing both the original and translated data makes the process of translation more transparent (ibid.). Th e process of transcribing is also complex and requires many decisions regarding the detail of the transcriptions and on how the data are presented to the reader. Another important, but often neglected issue is the translation of talk from one language to another. (ibid.) Th e data analysed here were translated from Finnish into English, which inevitably involves subtle loss of meaning. Also, the analysis was done on the original Finnish data, which were then translated into English. One source of diffi culty was the tendency of Finnish people to use the passive voice even when talking about their own doings, which translates very awkwardly into English. In cases where this led to my analyses confl icting with the English translation, I had no option but to exclude these sections from my reports. Another major decision that I had to make was whether or not to include the original Finnish data alongside the translated text, which adds to the transparency of analysis (Nikander 2008). I eventually decided against this, for two reasons.

Th e fi rst was a purely practical one, namely, the space restrictions placed on

journal articles. Another reason was a methodological one, but it concerned also the validity of analysis. For me it was more important to quote large data segments to give readers access to a wide variety of data and to illuminate the ways in which arguments were developed in interaction. Th is decision was a diffi cult compromise, and it can be considered a shortcoming of this research that the original data were not presented to the readers in the original articles.

On the other hand, this was done to improve readers’ access to the data and analysis.