Download Understanding Qualitative Research: Descriptive and Interpretive Approaches and more Lecture notes Qualitative research in PDF only on Docsity!
Chapter 11
Descriptive and interpretive
approaches to qualitative research
Robert Elliott and Ladislav Timulak
Qualitative research methods today are a diverse set, encompassing approaches such as empirical phenomenology, grounded theory, ethnography, protocol analysis and dis- course analysis. By one common definition (Polkinghorne, 1983), all these methods rely on linguistic rather than numerical data, and employ meaning-based rather than statistical forms of data analysis. Distinguishing between measuring things with words and measuring them in numbers, however, may not be a particularly useful way of characterising different approaches to research. Instead, other distinctive features of qualitative research may turn out to be of far greater importance (Elliott, 1999):
◆ emphasis on understanding phenomena in their own right (rather than from some outside perspective); ◆ open, exploratory research questions (vs. closed-ended hypotheses); ◆ unlimited, emergent description options (vs. predetermined choices or rating scales); ◆ (^) use of a special strategies for enhancing the credibility of design and analyses (see Elliott, Fischer and Rennie, 1999); and ◆ (^) definition of success conditions in terms of discovering something new (vs. confirming what was hypothesised). Space limitations preclude a complete survey of this rapidly growing field of research methods. Instead, we will focus on what are today regarded as established, well-used methods within the descriptive–interpretive branch of qualitative research. (In parti- cular, we will not cover discourse analysis, e.g., Potter and Wetherell, 1987; Stokoe and Wiggins, this volume, or ethnography.) These methods came into their own in the 1970s and 1980s, and have become mainstream in education, nursing, and increasingly in psychology, particularly in non-traditional or professional training schools. Because there is now an extensive history with these methods, standards of good research practice have emerged (e.g., Elliott et al ., 1999). Descriptive–interpretive qualitative research methods go by many ‘brand names’ in which various common elements are mixed and matched according to particular researchers’ predilections; currently popular variations include grounded theory
(Henwood and Pigeon, 1992; Strauss and Corbin, 1998), empirical phenomenology (Giorgi, 1975; Wertz, 1983), hermeneutic-interpretive research (Packer and Addison, 1989), interpretative phenomenological analysis (Smith, Jarman, Osborn, 1999), and Consensual Qualitative Research (Hill, Thompson, Williams, 1997). Following Barker, Pistrang and Elliott (2002), we find the emphasis on brand names to be confusing and somewhat proprietary. Thus, in our treatment here, we take a generic approach that emphasises common methodological practices rather than relatively minor differ- ences. It is our hope to be able to encourage readers to develop their own individual mix of methods that lend themselves to the topic under investigation and the researchers’ preferences and style of collecting and analysing qualitative data. In the approach to qualitative research we present here, we begin with the formu- lation of the research problem, followed by a discussion of issues in qualitative data collection and sampling. We will then go on to present common strategies of data analysis, before concluding by summarising principles of good practice in descriptive– interpretive qualitative research and providing suggestions for further reading and learning.
Formulating the problem
Previously, some qualitative researchers believed that it was better to go into the field without first reading the available literature. The reason for this position was the belief that becoming familiar with previous knowledge would ‘taint’ the researcher, predis- posing them to impose their preconceptions on the data and raising the danger of not being sensitive enough to allow the data speak for themselves in order to reveal essen- tial features of the phenomenon. It is our view that this approach is somewhat naive. For one thing, it is now understood that bias is an unavoidable part of the process of coming to know something and that knowledge is impossible without some kind of previous conceptual structure. Far from removing the researcher’s influence on the data, remaining ignorant of previous work on a phenomenon simply ensures that one’s work will be guided by uninformed rather than informed expectations. For this reason, the formulation of the research problem in qualitative research is similar in many ways to that in quantitative research. As a consequence, before com- mencing data collection, researchers carefully examine available knowledge and theory, carrying out a thorough literature search that includes up to date information on the topic of investigation. Strauss and Corbin (1998) refer to this as ‘theoretical sensitivity’ quoting Pasteur’s motto, ‘Discovery favours the prepared mind’. An important feature of this initial phase, however, is that the researcher should become as aware of possible of the nature of their pre-understandings of the phenom- enon, as these are likely to shape the data collection, analysis and interpretation. At the same time, the researcher should regard their expectations lightly, in a way that is open to unexpected meanings.
148 A HANDBOOK OF RESEARCH METHODS FOR CLINICAL AND HEALTH PSYCHOLOGY
an open-ended strategy for obtaining the data. By ‘open-ended’ we mean not only that participants are encouraged to elaborate on their accounts, or that observations are not restricted to certain pre-existing categories. Rather, open-endedness refers to the general strategy of data gathering. It means that inquiry is flexible and carefully adapted to the problem at hand and to the individual informant’s particular experi- ences and abilities to communicate those experiences, making each interview unique. There are several methods of obtaining information for qualitative inquiry. Qualitative interviews are the most common general approach, with semi-structured and unstructured interview formats predominating. In these forms of interview, participants are asked to provide elaborated accounts about particular experiences (e.g., tell me about a time when your asthma was particularly severe). The interviewer should have basic skills plus additional training in open-ended interviewing; such interviews are very similar to the empathic exploration found in good person-centred therapy (Mearns and Thorne, 1999). Good practice is to develop an interview guide that helps the interviewer focus the interview without imposing too much structure. Hill et al. (1997) recommend providing interviewees with a list of questions before the interview.
Variant formats
Self-report questionnaires are used much less in qualitative research, because they typically do not stimulate the needed level of elaboration sought by the qualitative researcher. However, given time and space constraints, questionnaires may be used as well. In that case they naturally consist of open-ended questions and ask respondents for elaboration, examples, etc. A good practice is to build in the opportunity to follow- up on questionnaires by phone interview (Hill et al ., 1997) or email correspondence, as responses often do not provide enough elaboration to understand the respondents’ point. A popular alternate form of qualitative interview is the focus group (see Wilkinson, this volume, for more information), a group format in which participants share and discuss their views of a particular topic (e.g., needed services for teenagers with spina bifida), allowing access to a large number of possible views and a replication of natur- alistic social influence and consensus processes. A special form of qualitative interview, used, for example, in research on helping processes (e.g., between breast cancer patients and their partners; Barker, Pistrang and Rutter, 1997), is tape-assisted recall. Here, a recording of an interaction is played back for the interviewee so that they can recall and describe their experience of particular moments (Elliott, 1986). Finally, think-aloud protocols (McLeod, 1999) are special forms of interview in which the participant is asked to verbalise their thought processes as they deal with a problem (e.g., managing an episode of high blood sugar). When observational methods are used in qualitative research they typically make exten- sive use of field notes or memos. These notes are primarily descriptive and observational
150 A HANDBOOK OF RESEARCH METHODS FOR CLINICAL AND HEALTH PSYCHOLOGY
but may also include the researcher’s interpretations and reactions, as long as these are clearly labelled as such. Qualitative observational methods often use non-interview archival data, such as tape recordings and associated transcripts of doctor–patient interactions. (Projective techniques can also be used as methods of data collection in qualitative research but are rarely seen, especially in clinical and health psychology.) There are three key aspects typical of the data collection in descriptive/interpretive qualitative research worth mentioning at this point: First, despite the fact that data collection in qualitative research generally does not use pre-existing categories for sorting the data, it always has a focus. The focus is nat- urally driven by the specific research questions. (At the same time, however, the general research approach encourages constructive critique and openness to reassessment of the chosen focus, if the data begin to point in a different direction.) Second, qualitative interviews are distinguished by their deliberate giving of power to respondents, in the sense that they become co-researchers. The interviewer tries to empower respondents to take the lead and to point out important features of the phe- nomenon as they see it. For example, respondents may be encouraged not only to reveal aspects of their experiences that were not expected by the researcher, but also to suggest improvements in the research procedure. Last but not least, a triangulation strategy is often used in this kind of research, with data gathered by multiple methods (e.g., observation and interviewing). This strategy can yield a richer and more balanced picture of the phenomenon, and also serves as a cross-validation method.
Specifics of sampling
Sampling in qualitative research, as in the quantitative tradition, is focused on the application of findings beyond the research sample. However, we cannot talk about generalisability in a traditional sense of stratified random sampling. Qualitative research does not aim at securing confidence intervals of studied variables around exact values in a population. Instead, qualitative research typically tries to sample broadly enough and to interview deeply enough that all the important aspects and variations of the studied phenomenon are captured in the sample – whether the sample be 8 or 100! Generalisability of specific population values or relationships is thus replaced by a thorough specification of the characteristics of the sample, so that one can make judgements about the applicability of the findings. As to the sample size, qualitative research does not use power analysis to determine the needed n , but instead mostly commonly uses the criterion of saturation (Strauss and Corbin, 1998), which means adding new cases to the point of diminishing returns, when no new information emerges. Obviously, given the nature of the data collected, and the time-consuming nature of analysis (see below) the size of the samples is usually much lower than in quantitative research.
DESCRIPTIVE AND INTERPRETIVE APPROACHES TO QUALITATIVE RESEARCH (^151)
voice in the data. During this stage of the analysis, it is worthwhile to read the whole data set, so that the researcher can get the whole picture of the studied phenomenon. During this initial reading, insights and understandings begin to emerge and are written down as memos. This is a kind of pre-analysis that can influence future steps of the analysis because the first relevancies start to unfold. During or after the initial reading an initial editing of the data often takes place. Obvious redundancies, repetitions, and unimportant digressions are omitted. One must, of course, be sure that the deleted data do not constitute important and relevant aspects of the phenomenon. Checks in the form of independent and challenging auditing processes by either the main analyst or others can be applied here also.
Delineating and processing meaning units
Next, we start to divide the data into distinctive meaning units (cf. Rennie, Phillips and Quartaro, 1988; Wertz, 1983). Meaning units are usually parts of the data that even if standing out of the context, would communicate sufficient information to provide a piece of meaning to the reader. The length of the meaning unit depends on the judge- ment of the researcher, who must assess how different lengths of meaning unit will affect the further steps of the analysis and who also should adopt a meaning unit size that is appropriate to their cognitive style and the data at hand. Generally, the longer the meaning unit is the bigger number (variety) of meanings it contains but the clearer its contextual meaning will be. As we delineate the meaning units, we can shorten them by getting rid of redund- ancies that do not change the meanings contained in them. For example, in a study of significant events in cognitive therapy with a diabetic patient, the data might read: ‘What was important for me was that the therapist verbalised exactly how I feel about my diabetes. The words she used helped me to be more aware of the things about it that I am having trouble with.’ A shortened version of this, which we would use in the further analysis, might then look as follows: ‘T verbalised exactly how P felt about her illness – it helped P to be more aware of what aspects of it P is having trouble with.’ (T stands for the therapist and P for the patient.) The meaning units are the units with which we do the analysis. However, it is good to be able to trace them back to the full data protocol, in case we need to be able clarify something from the context. For that reason it is a very good idea to assign a consec- utive code (in numbers and letters) to each meaning unit. The code should localise the unit in the original protocol. For example, if for each case we use different letter, we immediately know where meaning unit H82 came from. This procedure facilitates auditing.
Finding an overall organising structure for the data
Naturally, different sets of meaning units describe different aspects of the phenomenon. From their pre-research understanding of the phenomenon and their first reading of the data the researcher already has some ideas about some very broad headings for
DESCRIPTIVE AND INTERPRETIVE APPROACHES TO QUALITATIVE RESEARCH (^153)
organising the phenomenon into different processes or phases, referred to as domains. In fact, the researchers often introduce this structure into the data from the beginning via the interview question themselves. Nevertheless, the researcher typically waits until they finally sit down to code the data before developing a formal version of this organ- ising framework in relation to the first one or two data protocols. Consistent with this practice, Hill et al. (1997) recommend sorting the data into domains that provide a conceptual framework for the data, referred to in grounded theory as axial coding (Strauss and Corbin, 1998). As noted, this framework for meaningfully organising the data should be flexible and tested until it fits the data. It is important that the researcher always be open to using the data to restructure the organising conceptual framework. Critical auditing and testing out different possible frameworks are both useful strategies here. In studies reported in the literature, it is possible to find various kinds of relationship between domains, including temporal sequence (these things happened before these things), causes (this influenced this), significations (that is what this thing means now), etc. We typically find a variety of different types of structuring. Sometimes the domains may also mirror the different sources of the data, e.g., different kinds of observation, different kinds of self-report. If it is conceptually meaningful or in cases when the researcher is not sure about the structuring the data for the moment, it is possible to assign some data to more than one domain.
Generation of categories
Next, the meaning units are coded or categorised within each of the domains into which they have been organised. The categories evolve from the meanings in the meaning units. The word category refers to the aim of discerning regularities or similarities in the data (Glaser and Strauss, 1967). Creation of categories is an interpretive process on the part of the researcher (or in many cases the team of researchers, cf. Hill et al ., 1997), in which the researcher is trying to respect the data and use category labels close to the original language of participants. On the other hand, ideas for categories also come in part from the researcher’s knowledge of previous theorising and findings in other studies. Categorising is thus an interactive process in which priority is given to the data but understanding is inevitably facilitated by previous understanding. It is a kind of dialogue with the data. (In grounded theory this step is referred to as open coding .) The initial label for a category may come from the first occurrence of the meaning in some meaning unit; however, during the analysis it is typically refined as similar meanings are incorporated from subsequent meaning units. This evolving refining of categories means that some meaning units may eventually need to be reassigned to different categories. In this process, the meaning units are constantly compared (Glaser and Strauss, 1967) to each other and to the emerging categories, until all the data are sorted. At this point the analysis may contain thin or undeveloped categories or groups of categories lacking
154 A HANDBOOK OF RESEARCH METHODS FOR CLINICAL AND HEALTH PSYCHOLOGY
We can think of this part of the analysis as about the ultimate categorisation of the categorisation, or as about marketing what we found, or as about abstracting the results of our work. (In grounded theory this step is referred to as selective coding .)
Validity of analysis
The validity of the analysis is assessed throughout the study, as previously noted. To accomplish this, a constructively sceptical process of independent auditing is recom- mended. Although it is best for researchers to employ careful internal auditing throughout the analysis, the major auditing step typically occurs after a complete draft analysis has been produced. In addition, there are several other useful validation strategies. Validation by research participants is common, and involves presenting the results to the original informants or others like them in order to obtain feedback and correction. Another strategy, triangula- tion , involves comparing data collected by different methods (including quantitative). Collection of more cases may also be useful, particularly is auditing identifies problems with inadequate saturation of categories. Finally, resonation with the reader of the research paper is an essential form of validation in qualitative research. To facilitative this, the qualitative study should ground the findings in many illustrative examples, so readers may make their own judgements.
Interpretation of the results
Although categorisation of data uses interpretive strategies, one should not confuse this with the interpretation and discussion of the findings, which is done in qualitative research as in any other kind of research, after the findings are presented. As in quantita- tive research, it is important to place one’s qualitative findings within a context of previous theory and research findings. We can do this partly by reflecting thoroughly on the methodological influences and limitations shaping the results of our study (and previous studies). Probing the implications of our research can also include subjecting the existing literature to the same kind of rigorous process of analysis and categorisation as we used with our own data. In addition, it is also important to locate our findings within a socio-historical and scientific context, and to imagine useful further research.
Reviewing and critiquing qualitative research
In this chapter, we have tried to provide guidance on good practice in descriptive– interpretive qualitative research. Some of the key points of evaluation can be summed up in the following guidelines (see Elliott et al ., 1999, for more detail and examples of good and bad practice for each): 1 Own your perspective : Describe your theoretical orientation and personal interest in the research: values, interests, commitments, assumptions, expectations, and the role these played in the study.
156 A HANDBOOK OF RESEARCH METHODS FOR CLINICAL AND HEALTH PSYCHOLOGY
2 Describe your sample : Provide relevant features of the research participants and their life circumstances (e.g., age, gender, ethnicity, social class; variations in kind of experience described). 3 Ground categories in examples : Provide one or two concrete examples or each category (can be brief or extended). 4 Provide one or more credibility checks , including checking results with the original informants or others similar to them; using multiple qualitative analysts; using an additional analytic auditor (or a review by the original analyst); comparing two or more varied qualitative perspectives; or comparing results with quantitative data or external factors. 5 Organise categories to provide coherent understanding of how they fit together : Provide a data-based narrative, map, framework, or underlying structure to organise the phenomenon for the reader (e.g., memorably-named core categories or figures). 6 Accomplishing general vs. specific research tasks : For a general understanding of a phenomenon, use an appropriate sampling strategy and range of instances (informants or situations). For understanding a specific instance, make sure it has been studied and described systematically and comprehensively. 7 Allow readers to evaluate whether your categories resonate with their first- or second- hand experience of the phenomenon: Use concrete, rich language in order to help readers judge whether it has accurately represented the phenomenon.
Conclusions
Over the past 20 years, rigorous qualitative research methods have brought a breath of fresh air to the social sciences and increasingly to the health sciences, in particular re-habilitation and nursing. While our colleagues were at first highly suspicious of these approaches, they have subsequently learned that the kinds of methods we have been describing do not pose a threat to traditional quantitative methods but rather offer a useful complement for enriching, enlivening, and illuminating quantitative results. Nevertheless, these methods remain generally under-utilised by psychologists, including health psychologists. We urge readers to take the plunge by adopting a more pluralist approach to research incorporating the methods described here along with other, more traditional approaches.
Further reading
Miles and Huberman (1994), Denzin and Lincoln (2000), and Creswell (1998) provide varied but useful general treatments each covering a range of approaches to qualitative research. The central sources for the particular version of qualitative research described in this chapter are Rennie et al. (1998) and Wertz (1983); see also Strauss and Corbin (1998) on grounded theory; Hill et al. (1997) on Consensual Qualitative Research, and
DESCRIPTIVE AND INTERPRETIVE APPROACHES TO QUALITATIVE RESEARCH (^157)
Pistrang, N., Barker, C. and Rutter, C. (1997). Social support as conversation: Analysing breast cancer patients’ interactions with their partners. Social Science and Medicine , 45 , 773–782. Polkinghorne, D. (1983). Methodology for the human sciences. Albany, NY: Human Sciences Press. Potter, J. and Wetherell, M. (1987). Discourse and social psychology. London: Sage. Rennie, D. (1990). Toward a representation of the client’s experience of the psychotherapy hour. In G. Lietaer, J. Rombauts, and R. Van Balen (eds) Client-centered and experiential psychotherapy in the nineties , pp. 155–172. Leuven, Belgium: Leuven University Press. Rennie, D. L., Phillips, J. R., Quartaro, G. K. (1988). Grounded theory: A promising approach to conceptualization in psychology? Canadian Psychology , 29 , 139–150. Smith, J. A., Jarman, M. and Osborn, M. (1999). Doing interpretative phenomenological analysis. In M. Murray and K. Chamberlain (eds) Qualitative health psychology , pp. 218–240. London: Sage. Strauss, A. and Corbin, J. (1998). Basics of qualitative research: Techniques and procedures for developing grounded theory , 2nd edn. Newbury Park, CA: Sage. Timulak, L. and Elliott, R. (2003). Empowerment events in process-experiential psychotherapy of depression. Psychotherapy Research , 13 , 443–460. Wertz, F. J. (1983). From everyday to psychological description: Analyzing the moments of a qualitative data analysis. Journal of Phenomenological Psychology , 14 , 197–241.
DESCRIPTIVE AND INTERPRETIVE APPROACHES TO QUALITATIVE RESEARCH (^159)
