RESEARCHING THE REAL WORLD



MAIN MENU

Basics

Orientation Observation In-depth interviews Document analysis and semiology Conversation and discourse analysis Secondary Data Surveys Experiments Ethics Research outcomes
Conclusion

References

Activities

Social Research Glossary

About Researching the Real World

Search

Contact

© Lee Harvey 2012–2018

Page updated 16 November, 2018

Citation reference: Harvey, L., 2012–2018, Researching the Real World, available at qualityresearchinternational.com/methodology
All rights belong to author.


 

A Guide to Methodology

1. Basics

Activity 1.15.1

1.15 Triangulation
All research methods have advantages, disadvantages and limitations. Research methods are not neutral in how they represent the world. Some researchers have therefore used more than one approach as a means of dealing with this problem in a study. This process is called triangulation. Triangulation can apply to methods, theory and epistemology.

In its broadest sense, triangulation refers to a combination of ways of exploring a research question. Triangulation can be broken down into four types:

1. researcher (also referred to as investigator triangulation); (1.15.1)
2. theory; (1.15.2)
3. method (or data); (1.15.3)
4. methodological. (1.15.4)

Technical note: Denzin (1970) is often attributed with identifying four types of triangulation but his four were the first three above with method and data split up to create four. However, to add confusion, he called 'method' triangulation 'methodological' triangulation but he meant the use of more than one method, rather than the combining of different methodological approaches.

In addition, this section includes:
an alternative categorisation of triangulation (1.15.5)
and a note on meta-analysis (1.15.6)

Top

1.15.1 Researcher triangulation
Data can be triangulated by comparing the data collected by one researcher with that collected by other researchers using the same methods with different sections of the same sample. In its simplest form, this might be a check on the accuracy of data collected by different members of a team of interviewers.

A rather more complex situation occurs when a team of researchers is undertaking participant observation on the same topic in different locations. Different data may be the result of differences in approach by the researchers or genuine differences in the research setting that each is observing.

Community studies provide an example of triangulation where two or more researchers use the same technique. A team of researchers undertake participant observation of the community from different points of view. They adopt different roles in the community, which provides them with more than one perspective on the research situation. An example of this is Davis, Gardner and Gardner's (1941) classic American study of the relationships between blacks and whites in an old southern city. Four researchers, a black husband and wife and a white husband and wife, lived in the city for almost two years. They had very different views of the community due to their different roles and positions. Triangulation came about through the collection and comparison of their different perceptions of the community.

Similarly, in Power, Persistence and Change: A Second Study of Banbury (Stacey et al., 1975) Eric Batstone, Anne Murcott and Colin Bell all moved to Banbury for the duration of the two-year fieldwork and bought houses there. They were participant observers in the town and also carried out in-depth interviews with key informants as well as a social survey of a sample of 1,449 residents.

It is argued that using teams of researchers provides a check on peoples' views, perceptions and interpretations of the same situation and thus overcomes the 'subjectivity' of the research process. This is a particularly popular view concerning participant observation, which is often seen as 'rather subjective' and it is assumed that having several people involved overcomes the problem of idiosyncratic individual interpretations. This, however, does not mean that the consensus view is any more 'correct' or objective, as no method is intrinsically objective.

Top

1.15.2 Theory triangulation
Theory triangulation is the analysis and comparison of two or more theoretical positions relating to the research problem. In effect, researchers normally undertake a degree of theory triangulation as a matter of course when undertaking a literature review prior to collecting new empirical data. The theory triangulation informs the research plan.

Theory triangulation thus involves looking at the research situation from different theoretical perspectives. For example, a study of juvenile gangs might adopt both a functionalist and an interactionist perspective. The assumption is that by applying more than one theoretical approach, a better understanding might evolve.

An example of triangulation of theory is Hochschild's (1983) study of airline flight attendants, which is based on multiple research strategies and data sources. In the analysis, she moves between interactionists theories of emotion and Marxian theories of political economy to bring together micro analyses of face-to-face interaction with macro analyses of power, economic and social organisation.

Theory triangulation may be more formalised: the research design may set out to test competing theories. For example, Singer (1971) used three different theories in his research on why first-born children were more adult-oriented than later-born children. He devised a research strategy to test each of the competing theories.

However, theory triangulation may also be used subsequently to help make sense of data that do not seem to corroborate or relate to any individual theory.

Top

1.15.3 Method triangulation
The different methods discussed in this book are capable of being used on their own. However, the decision to make use of triangulation is sometimes based on the belief of some researchers that no single method will provide a complete picture of the area under study. The use of different methods might occur at different stages of the research or they may be used simultaneously.

For example, a researcher may use focus groups and in-depth interviews to identify the issues relevant to a group of people that will later be surveyed using a structured interview. The qualitative methods are employed at the exploratory stage and the quantitative methods are used to provide statistical indicators and to suggest generalisable relationships between types of respondents and types of answer.

For example, Colin Lacey (1970) used both observation and scheduled interviews in his education research. He first undertook qualitative case studies to arrive at a model and then used quantitative analysis on a bigger sample to try to verify his qualitative model. Similarly, Bosk (1979) used a variety of techniques in his study of surgeons in a teaching hospital (CASE STUDY Bosk triangulation).

Activity 1.15.1
Read CASE STUDY Bosk triangulation
1. What methods of data collection does Bosk use?
2. What sort of gaps in his observational data do the other techniques fill?
3. What type of interviewing does he adopt?
4. He argues that the use of several methods makes his inferences valid. Do you agree?
5. What might you do in his situation when conclusions from one source of data are not confirmed by another source of data?.
.
As a class activity this would take about 15 minutes in small groups with a short 5-minute feedback session.

Alternatively, a researcher may, for example, be exploring drug abuse in a small town and might use police records, statistics on convictions, local newspaper reports and non-participant observation all as sources of information to explore the extent and nature of drug abuse.

In this simultaneous use of different methods, the researcher has to consider whether the different methods provide different means to getting at 'the data' (also sometimes called 'data-triangulation') or whether the different methods are indicative of different methodological approaches that need to be reconciled (methodological triangulation, see below).

Data triangulation is the use of two or more sets of data derived from the same method or different methods. For example, the level of unemployment in a geographical area could, in the UK, be obtained from the Labour Force Survey conducted by the Office of Population Census and Surveys, from unemployment statistics based on the registered unemployed, from statistics estimated by the Low Pay Unit or by a local survey. Each of these figures would be different and, taking into account the limitations of each, it would be possible to triangulate to derive a figure that was as accurate as possible. It also allows the researcher to evaluate the sources by comparing the data collected.

However, the perception of triangulation as 'data triangulation', that is, that measuring the same research problem from different angles provides you with a better reading or measurement of it, can be a problem. Different data sources are likely to provide information derived from different epistemological or ontological perspectives and thus, implicitly, are data relating to different research questions. 'Data triangulation' also implies a view of the social world in which there is one objective and knowable social reality and all that social researchers have to do is to work out which are the most appropriate triangulation points by which to measure it.

Throughout this Guide we argue that there are various perspectives that researchers employ and have questioned the notion of there being a single objective social world.

In practice, method triangulation goes beyond 'data triangulation', that is it goes beyond simply using different methods to get at the same thing. More likely, method triangulation involves adopting different methods to look at a research problem from different perspectives, which can be used by the researcher as a means of comparison and contrast.

For example, a case study focusing on drug taking in prison may involve several methods. The researcher could decide to use a questionnaire with staff and prisoners or the researcher could choose to use a variety of methods to collect the data such as, a focus group with prisoners, a questionnaire to all the prison staff and some in-depth interviews with prison officers. Using multi-methods produces different kinds of data on the same topic. The use of a questionnaire to all staff generated one kind of data and the in-depth interviews and focus groups generated 'rich' (that is, detailed) data by allowing the subjects of the research to raise the points of importance to them.

Furthermore, in this example, constraints of time, resources and access necessitated the use of different methods. For example, using focus groups with prisoners allowed a much wider group to be interviewed than would have been possible if in-depth interviews had been used. In addition, prisons have very regimented schedules and neither the researcher nor the prisoners are able to move around the prison freely. Therefore, it was easier to arrange to have a group of prisoners together in a focus group than to arrange individual interviews (MacDonald, 1998).

Eileen Moyer et al. (2013), for example, used a range of ethnographic methods, including participant observation, key informant interviews, in-depth interviews and focus group discussions in their study of HIV disclosure in Kenyan health facilities.

Using a range of methods may also allow the findings from one method to be checked against the findings from another. The use of multi-methods allows findings to be corroborated or questioned by comparing the data produced by the different methods. In the prison research example, the data from the prison officer questionnaires corroborated the data from the prisoner focus groups about the use of drugs in the prison and conflicted with official views about the nature and extent of drug use in prisons.

In her study of the lives of working women in Austria and Hungary, Fodor (2003) used a mixture of statistical analysis, in-depth interviewing and archive research. She triangulated the material as follows. She undertook regression analysis of data from the 1980s to reveal the factors that had an impact on the access of women, in both countries, to élite managerial positions in the economy or government. She then constructed models to show the difference in the two countries. The bare statistical analysis was given life by her depth interviews with female managers. This was further enhanced by material from the archives of the Hungarian Communist Party and the Austrian Parliament.

To sum up, we could argue that researchers may find it helpful to use more than one method to corroborate or question their data. Researchers should also be aware that mehod triangulation is not a simple process and findings that come from the use of multi-methods have to be treated cautiously. Furthermore, researchers should not consider that by using triangulation their data is somehow 'proven' or considered to be absolutely correct. Using method triangulation can often raise more questions than it answers.

Top

1.15.4 Methodological triangulation
Methodological triangulation is the most complex issue to resolve. It brings together issues of different techniques, data sources and perspectives of subjects and researchers. Essentially, it requires exploring the extent to which different elements of the data are dependent upon different epistemological presuppositions. Information collected via a questionnaire might be premised on the implicit assumption that it shows broad trends and correlations based on the researcher's perspective of the research problem—an essentially positivist approach. This might be alongside data collected via in-depth interviews that are seeking to get at the meanings of the subjects—a phenomenological approach. How, then are these to be reconciled?

Methodological triangulation is not straightforward in that there is no simple way to resolve epistemological differences.

The problem cannot be overcome by assuming that one particular broad method, such as interviewing, is indicative of a particular epistemological framework. For example, the status of the data achieved from an interview is different depending on your perspective. As part of a positivistic approach the interview is seen as a way to gather 'facts' about behaviour or attitudes. (It is likely in this case that the researcher will ensure standardisation in the interviewing process normally using structured interviews with a randomly selected sample.) In phenomenological or critical social research, the interview gives rise to rich data exploring the interviewees interpretations or understandings, focusing on what the respondents see as the crucial issues. (This type of interview would normally be unstructured or semi-structured and subject to reflexivity rather than verification.) Thus the data produced from the above two interviews may well be very different or even contradictory and it then becomes problematic to decide which set of data takes precedence.

Tiainen and  Koivunen (2006, pp. 7–8) provided an example of multiple forms of triangulation. They explained how, in their study of the human and technological dynamics around some newly offered electronic services in a specific rural area of Finland, they used multiple researchers, data sources, methods and theories. They considered it avoided imposing any particular interpretation on the research process, which they described as follows.

Multiple data sources
In our study of the villagers' ICT use, data were gathered from public documents, interviews, participatory observations, and trial use of technical artefacts. Public documents included the following: official statistics, municipal annual reports, local newspapers, and village Internet sites. About 60 open-ended interviews, conducted in 2003 and 2004, were a major source of data. We first interviewed active villagers (e.g., members of the village residents' association) and the owners of local enterprises. We also asked them about other possible informants whom we contacted later.
Participant observation was made and field notes were taken by one researcher who participated in some local activities, such as village events and voluntary work in a computer workshop. Moreover, almost all the interviews happened in informants' homes, which gave us a chance to see the informants' living contexts.

Multiple Methods
...we made use of several data gathering methods--broadly speaking, participant observation and ethnographic interviews. We also collected some quantitative data, primarily as background information on the village.
For analysing the data, we used discourse analysis and phenomenography. ...these methods were sufficiently different from each other to serve as independent qualifying perspectives, fulfilling the aim of triangulation.

Multiple Theories
Our research group is located in a department of computer sciences and the group leader is a professor of information systems [in] information systems ...people are seen in relation to technology; they are called users …. Another group member had a background in consumer studies. This field affords a wider view of human beings by focusing on their experiences, motivations, behaviours, and attitudes....Yet another member had a background in social anthropology. Social anthropology is concerned with knowledge about humans in societies. The main focus is on the diversity of social life....Thus, we tried several theoretical concepts to grapple with the situation in the field. We analysed people's role in technological developments by using the frameworks of social shaping of technology (Bijker, 1995), diffusion of innovations ( Rogers, 1995), and ICT domestication (Lie & Sorenson, 1996; Silverstone & Hirsch, 1992). The different theoretical concepts helped us appreciate the situation in the field from different perspectives. This also led to interesting results. For example, we produced a revised version of the domestication theory, augmenting it with the concepts of community and social practice, taking these latter elements from the theory of diffusion of innovations.

Multiple Researchers
In our project, we had several researchers doing the same task. We did this in both data gathering and data analysis....We made contact with the villagers in the meetings of the residents' association, using the snowball sampling method....We conducted some collaborative interviews. These included two researchers interviewing one informant together and two researchers interviewing different members of a family at the same time. In the designing the interviews, we tried to decrease the researcher bias...

In conclusion, as Olsen (2004) argued, triangulation is not just about validation but it also aims 'at deepening and widening one's understanding'. In that sense she would dispute O'Donoghue and Punch's (2003, p. 78) view that triangulation is 'method of cross-checking data from multiple sources to search for regularities in the research data'.

Study Point
Does the use of triangulation necessarily improve research outcomes?

Top

1.15.5 An alternative analysis
In an article entitled 'Troubles with Triangulation', Hammersley 2008 proposed an alternative way of classifying triangulation.

The first, is triangulation as validity-checking, which reflects the approach mentioned above by O'Donoghue and Punch (2003). For Hammersley (2008, p. 23), this involves:

checking the validity of an interpretation based on a single source of data by recourse to at least one further source that is of a strategically different type...this does not necessarily involve combining different methods of data collection: for instance, it might require comparing interview data from several witnesses to an event...or it could involve comparing observational data from various settings that bear on the same knowledge claim....The idea behind this first concept of triangulation is that by drawing data from sources that have very different potential threats to validity it is possible reduce the chances of reaching false conclusions.

The second, is indefinite triangulation, which Hammersely, drawing on Cicourel 1974, decribes as not an attempt to assess accounts to identify the truth about a specific situation:

Rather, the approach adopted is closer to the sociology of knowledge: the interest is in why participants' accounts take the varying forms they do, or rather in how they have been put together....Built into the ethnomethodological position drawn on here is a denial that there can be only one true statement about relevant features of the situation to which various accounts relate, and (even more significantly) a rejection of the idea that social science can or should adjudicate amongst informants' accounts.

Hammersely (2008, p. 25) argues that indefinite triangulation is 'a device for generating divergent interpretations, rather than for checking the validity of inferences from data'.

The third category is triangulation as seeking complementary information, which Hammersely decribes, drawing on Erzberger and Kelle (2003), as the use of different methods to provide two or more distict viewpoints or angles. These viewpoints are not for validation but are intended to provide a fuller and more complete picture of the phenomenon concerned. Hence the idea of complementarity. Hammersley (2008, p. 26) suggests that this is 'perhaps today the most common meaning of the term routinely employed by researchers'.

The fourth category is triangulation as epistemological dialogue or juxtaposition, which Hammersley describes, citing Flick (1998) as follows:

Flick argues that different methods do not simply provide varying kinds of information about the same object, but constitute the world in different ways....[Thus] different methods construct the social world in divergent ways, so that combining them may not lead either to validation or to increasing the completeness of the picture. (Hammersley, 2008, p. 27)

There is a problem with this approach, Hammersely argues, in that it potentially destroys triangulation because it fails to resolve any of the tensions or different accounts that come from different epistemological positions.

This position perhaps reflects a refusal to choose among epistemological paradigms, or to let the reader do this easily. Instead, the goal is to put, and to keep, methods and epistemologies both in tension and in question, along with throwing doubt on any idea that one or other approach is correct, or that the differences between them can be overcome. We might call this 'postmodernist triangulation'. (Hammersely, 2008, p. 27)

This in a way this reflects methodological triangulation discussed above but rather than take the postmodernist way out and fail to resolve completing perspectives, a critical approach to methodological trinagulation would use the tensions or contradictions as a basis for deconstruction and subsequent reconstruction of an alternative understanding (see section 2.4)

Top

1.15.6 Triangulation and meta-analysis
Triangulation is different from meta-analysis. Triangulation is a 'within' study technique while meta-analysis involves the pulling together of already-completed research studies. A critical literature review, for example, is a form of meta-analysis. Some, though would argue that meta-analysis requires the research studies to be similar and 'combinable' (and implicitly quantitative). For example:

Meta-analysis combines the original data from several rigorous scientific studies of similar quality and design for sophisticated statistical analysis. In contrast, triangulation uses findings from diverse sources, bearing in mind the strengths and weaknesses of those findings, and it looks for a convergence of the evidence in order to draw overall conclusions. (UNAIDS, 2010)

Top

1.15.7 Advantages and disadvantages of triangulation
Triangulation is often assumed to be beneficial in any research situation. The benefits include:

1. The ability to cross-check perceptions when a team of researchers is used.
2. Cross-checking research findings by using more than one data collection method.
3. Checking out different theories using the same set of data.

However, there are also some problems with triangulation. First, there is the problem of available resources. Using more than one method can be time-consuming and expensive. It may involve employing both interviewers and observers.

Adopting theoretical triangulation can be less resource-demanding although it may lead to the researcher needing to undertake more research to validate or assess the theoretical perspectives.

Second, using a team of people can create tensions when it comes to decision-making, sharing of work and supervision. Colin Bell (1977, pp. 54­–5) reflecting on the Banbury study, recounted the problem of teamwork, particularly the friction that occurred over the use of the interview survey.

The survey itself was a lot of work, cost a lot of money (it was the biggest item on the budget after our salaries) and was very time-consuming. And as the bitter recriminations later showed, it was extraordinarily difficult to do well.... In the summer of 1968 Anne Murcott and Eric Batstone went to Swansea and I went to Colchester. By that time we had all had more than enough both of each other and of Banbury.

Third, when using triangulation there is a tendency to adopt the 'safe view' in an attempt to reach agreement. Thus potentially critical research material can be overlooked or left out altogether.

Fourth, triangulation raises epistemological problems. For example, combining data that has been collected via participant observation with social survey results can be difficult. The observation may have been phenomenological, looking for subjects' meanings while the survey results may have involved a positivistic search for causes. Often the two methods do not balance and one approach overrides the other. Lacey (1970), as noted above (Section 1.15.3), used his quantitative data to verify what he had learned through observation. The question remains: what do does the researcher do in  a case where the observations and the survey data suggest very different interpretations? Do you decide that the survey material is more 'objective' and use that? Do you admit to having contradictory evidence and say that you do not know what your data tells you? The answer is to make it clear, from the outset, what the underlying presuppositions are. In the last resort, it is these presuppositions that will guide the researcher in dealing with contradictory types of evidence. The researcher should not be swayed by the data collection method but should consider the evidence in light of the underlying perspectives and the aim of the research.

Next 2 Orientation