7

Issues in Qualitative Data Collection

Introduction

The first question when deciding on a data collection method is this: ‘What sort of data will help me to answer my research question or achieve my research aims?’ The second question is then: ‘What is the most appropriate method for collecting those data?’ This chapter considers issues related to collecting qualitative data.

As we noted in Chapter 3, it is always the research question that determines the method. There is a danger, sometimes, in researchers limiting themselves, when writing their research questions, by having half an eye on the methods they want to use. If we want to be honest and clear in doing research, we need, first, to formulate our question very clearly and only then decide on method. For the purposes of the rest of the chapter, it is supposed that the researcher has done this and has decided that qualitative data are needed. As we have seen in other chapters, the form of these data is usually textual. Qualitative research is normally about the generation and analysis of words. Those words can take at least three forms: words uttered in interviews, words written down by the researcher following or during observation or words that are already published. Following an examination of sampling issues which might affect the collection of these forms of words, we will look at how each different form of words might be gathered.

Sampling for qualitative research

It should be noted that whilst an aim in quantitative research is usually to generalise from the findings (and this is often possible because of the nature of the sampling used), the aim of qualitative research is not to generalise (because the nature of the sampling and the methods of data collection and analysis do not allow it). This is an important point that must always be borne in mind when reporting qualitative findings. Nevertheless, the findings from qualitative samples may be illustrative of particular experiences and points of view. Indeed, the kind of sampling and data collection strategies employed by qualitative researchers often allow very detailed and deep descriptions of personal accounts.

In Chapter 2, we touched on three popular forms of sampling for qualitative research: convenience sampling, purposive sampling and snowball sampling. We noted that convenience sampling involved sampling from people who happened to be available to participate, whilst purposive sampling involved selecting people on the basis of their being likely to have things to say relevant to the research aim, and snowball sampling was a particular variant of purposive sampling involving having respondents themselves recommend potential additional respondents. None of these approaches to sampling attempt to be representative. However, as we saw in Chapter 6, this is not necessarily a drawback.

Whilst convenience sampling perhaps has least to recommend it, it still has distinct advantages, not least of which is that participants are readily available. This may be a compelling advantage for hard pressed students with limited resources. The great disadvantage, however, is that members of a convenience sample may have little to say of use in the illumination of the research question. For example, an examination of the experiences of social workers of working with people with drug problems might run into this issue of relevance if our convenience sample contained only social workers who had limited experience of caring for such clients. In a quantitative study, we might be able to make use of their limited responses, but for qualitative research, it is unlikely we would get the kind of detailed personal accounts which are at the heart of the qualitative approach to research. On the other hand, almost any convenience sample of social workers could give us such detailed accounts of their experiences of having been students. In such a situation, the weaknesses of a convenience sample would be trivial compared with the great advantage of availability.

Where the phenomena we want to study are fairly common, then, a convenience approach has a lot to recommend it. However, the more unusual the phenomenon, the more likely it is that we will want to turn to purposive sampling as a way of recruiting people with relevant information. In purposive sampling, the researcher sets out a specific set of criteria according to which participants will be selected, and then recruits as many participants as are required who meet these criteria. Often, purposive sampling is highly selective, resulting in a unique group of individuals, and it may be that all available participants are recruited. Patton (1990) has suggested a considerable number of differing purposes for purposive sampling, dependent on the differing possible aims of the researcher. For example, during a project, the search for deviant cases (who point to unusual aspect of the issue under study), typical cases (who show typical examples of the issue), theory-based cases (who demonstrate characteristics in line with a particular theory and so allow us to expand upon that theory) are all types of purposive searching for participants. About 20 other types of purposive sampling are identified in Patton’s book.

Sometimes, it is difficult to find a way into a particular cultural group (drug users, abused women, criminals). Once again, a form of purposive sampling is a useful tool for the qualitative researcher. In this case, snowball sampling is the approach of choice. It relies on the idea that people in particular communities are best placed to direct the researcher to potential sources of further information. The more closed or covert the community, the more likely it is that the researcher will need to make use of the snowball approach, but at the same time, the more difficult this will become, because the research will need to work hard to gain the trust of initial respondents, so that they will feel comfortable in commending the researcher to other potential participants. However, once that trust is gained, the snowball approach is a powerful way of gaining information from hard-to-reach participant groups. There is, however, an ethical dilemma in some instances of the use of snowballing with hard-to-reach groups. This is the possibility that harm may come to a respondent as a result of putting the researcher in contact with other potential participants. The case of criminal behaviour is an obvious example. A potential respondent may be less than happy about having their status as one who engages in illegal activities revealed to a third party in the form of a researcher whose pedigree and motivations are unknown to them.

It is impossible to give an immediate illustration of a typical size of a sample in quantitative research. Typical sizes of samples in qualitative research range from a single case to about 30 respondents. The literature offers little conclusive evidence of the importance of particular sample sizes in qualitative research. One method that is sometimes used to limit samples in qualitative studies is to continue to interview respondents, or collect other forms of data, until the same information starts being repeated over and over again with no new information emerging. This is sometimes referred to as ‘saturation’. Interviews or other forms of data collection can stop at this point and thus the sample size is limited by the nature of the forthcoming data. Naturally, it takes considerable skill and experience to reach a reasonable decision as to when no new information is likely to emerge.

Could qualitative researchers use randomised samples?

The standard view on qualitative sampling is that it is not necessary to use a random sample. Qualitative research, so the argument goes, reflects personal, subjective reporting of personal, subjective data. However, it is possible to question this point of view.

One of the reasons that qualitative findings cannot be generalised is because of the sampling methods used. In the past, qualitative researchers had to do most things by hand. They wrote down interviews and then analysed them – also by hand. Today, there are many technologies that help make the data collection and data analysis processes easier. Clearly, interviews can be recorded. They can also be analysed with the aid of computer programs. Such programs do not do the analysis but are a potent aid to the process. Such programs can work with computer-based transcripts of interviews or even with the direct voicerecordings of those interviews.

All this raises the tentative question as to whether or not large-scale qualitative studies, involving randomised samples might be conducted, from which the findings could be generalised. Some view this idea as heresay or as a plain misunderstanding of the nature of qualitative research. Like most things, however, it is worth thinking about.

Interviews and focus groups

In qualitative research, three types of interviews can be noted: structure, semi-structured and unstructured. Along that continuum comes an increasing difficulty in later analysis. Structured interviews are probably the easiest to analyse (or, perhaps, to ‘organise’). Semi-structured interviews are rather more difficult and unstructured interviews can become so wayward that they can be very difficult – if not impossible – to analyse. However, in terms of value and usefulness, the gradient seems, sometimes, to be reversed: unstructured interviews can sometimes lead us closer to other people’s meaning systems, whilst structured ones can be very arid affairs. Another method of data collection is the ‘group interview’ or focus group.

The structured interview

This form of interview can be likened to a sort of ‘verbal questionnaire’. The research prepares a list of questions which the interviewee then answers. No leeway is given for alternative questions or for any further ones. It is usually important that the interview schedules that are prepared for these interviews contain, almost exclusively, open questions (as opposed to closed questions). A closed question is one that elicits only a one word answer – and very often, the answer is either ‘yes’ or ‘no’. On the other hand, an open question is broader and asks the interviewee to express a view or an opinion. If only closed questions are to be asked, then the research needs to consider if it would not be more economical of time and energy to use a questionnaire rather than an interview. Moreover, the data gathered from such closed questions is hardly in keeping with the spirit of qualitative research.

The big advantage of the structured interview is in the marshalling of the data, afterwards. The responses to each question, from each interviewee, can be brought together and reported. A particularly straightforward way of doing this is, for reporting purposes, to turn the question used in the interview, into a statement. Thus, if the question was ‘What are your views about the value of reflective practice in the clinical setting?’, it is turned, in the report, into the heading: ‘Student’s views of the value of reflective practice in the clinical setting’.

The big disadvantage is that the structured interview does not allow the interview room to encourage the interviewee to expand on what they have to say. The researcher is limited to only asking the questions as they appear on the interview schedule. The structured interview is useful, in a large-scale project, if a number of interviewers (apart from the researcher) are to be trained to collect data. The structure ensures that data of a similar sort will routinely be collected from each respondent – despite the fact that different interviewers are being used. This approach is often used for large-scale surveys and opinion polls.

Semi-structured interviews

This is a variant on the above approach. The researcher who uses the semi-structured approach has a series of questions that they wish to ask. Or, they may have a series of topics in mind, around which they will frame questions. However, the researcher is also prepared to ask subsequent questions and to use prompts to encourage the interviewee to say more. A decision may also be made not necessarily to ask the questions in a particular order but to change the order according to the respondents’ responses. This approach encourages the interview to flow more freely: one topic often leads, seamlessly, into another until the interviewer has covered all her questions. The general aim is to cover similar territory, in each interview, but not to control the interview to the degree that each respondent is asked exactly the same questions in exactly the same order.

The researcher needs to practice this method before she uses it and some counselling skills training can sometimes be helpful in encouraging the researcher to use prompts and other devices to help the respondent to keep talking. However, there is an important difference in the aim of counselling and data collection. While counselling is mostly about helping a person to talk about their problems and helping them sort out those problems, the aim of the research interview is to collect data. The research interview is never, in any sense, a form of counselling or therapy. Indeed, if the issues under discussion are sensitive, the researcher should have contingency plans to cover what happens if the respondent becomes upset. It is not normally ethically justifiable to continue to gather data when a person is upset, nor may it be justifiable to use the data so collected.

Stay updated, free articles. Join our Telegram channel

Join