We're currently redeveloping our website and you may notice some inconsistencies with our designs.

Triangulation

main content

Triangulation

Introduction

This topic is concerned with triangulation. Miles and Huberman (1984: 234) have a nice way of explaining it:

Detectives, car mechanics and general practitioners all engage successfully in establishing and corroborating findings with little elaborate instrumentation. They often use a modus operandi approach, which consists largely of triangulating independent indices. When the detective amasses fingerprints, hair samples, alibis, eyewitness accounts and the like, a case is being made that presumably fits one suspect far better than others. Diagnosis of engine failure or chest pain follows a similar pattern. All the signs presumably point to the same conclusion. Note the importance of having different kinds of measurement, which provide repeated verification.

Anyone who has experienced recurrent problems with a car the garage has been unable to fix, or read in the press about miscarriages of justice over recent years, may not share Miles and Huberman’s confidence in the ability of car mechanics and detectives to reach the ‘right’ conclusion! Putting that to one side, however, their analogy is a useful one. It shows that a variety of strategies is needed to fully understand and explain a problem: the challenge is to find the right combination of strategies and then bring together the findings, weighing up their cumulative evidence.

In this topic we begin by looking at: the origins of and rationale for triangulation. We then consider the four main ways it can be achieved:

  • triangulating methods,
  • triangulation investigators or
  • triangulation of data (sometimes called the ‘realist’ dimensions of triangulation),
  • triangulation of theory.

We end on a cautionary note.

Origins and Rationale

The concept of triangulation, originating from ancient Greek mathematics, is applied in a wide range of spheres, including geometry and surveying. In navigation, it is used to establish a ship’s position: measuring a vessel’s distance from more than one point on the shore gives a more accurate reading of its location. Triangulation was first applied to research by Campbell and Fiske (1959) and developed by Webb (1966), who argued that researchers should employ more than one instrument to measure variables. As this implies, triangulation was first associated with quantitative research but its relevance to qualitative methods was soon explored. Denzin (1970, 1978) was a major proponent of the use of triangulation by researchers working within the interpretivist paradigm.

Some of the arguments for triangulation have already been discussed in Topics 1 and 2 and will not be repeated here. Denzin’s rationale stresses the interdependence of theory and method. Writing from a symbolic interactionist perspective, he views the researcher as an active participant in the social world, which he or she ‘acts on’ through the use of method.  Denzin asks the reader to think of two researchers studying a psychiatric hospital. Each chooses different methods: one opts for a survey while the other uses participant observation. This leads to differences in the questions they ask and the observations they make. In addition, the findings are coloured by the researchers’ different personalities, biographies and biases, all of which influence the nature of their interactions with the social world. Each uncovers different aspects of what takes place in the hospital but neither can reveal all of it. Therefore, Denzin concludes, to get as full and as accurate a picture as possible, researchers must use more than one strategy. 

Triangulation Of Methods

Denzin identifies two types of methodological triangulation – ‘within-method’ and ‘between’ or ‘across’ method. A study using the ‘within-method’ approach is confined to one method but uses different strategies within it. Denzin is rather dismissive of this approach, giving the example of a survey which uses a number of different scales to measure the same variable. The researcher may think he is measuring the same phenomenon in several ways but the decision to stay within one method means the biases associated with that method remain unchallenged.

Much more satisfactory, in Denzin’s opinion, is the ‘between’ or ‘across methods’ approach which, by combining at least two different methods in one study, reaps the benefits of each approach while also compensating for their weaknesses, an argument we met in Topic 2. Denzin (1978: 303) identifies four principles of methodological triangulation, namely:

  • The nature of the research problem and its relevance to a particular method should be assessed and, where necessary, the method tailored to the problem at hand
  • Methods should be combined with a ‘checks and balances’ approach so that threats to internal and external validity are reduced as much as possible; ie: the particular weakness of one method is compensated for by the particular strength of another
  • The theoretical relevance of each method must be considered as well as the implications of combining methods which at first may appear contradictory
  • Researchers should continually reflect on their methods, being ready to develop or alter them in the light of developments in the field and emerging data.

Have a look at an interesting study by Angela Johnson (2007) which used methodological triangulation to explore the experiences of women science undergraduates from three Black and minority ethnic groups. As well as interviewing the students, Johnson attended and observed their science classes. She coded the two data sets separately and then compared them for common themes. This enabled her to show how cultural values and teaching practices within science departments caused difficulties for these particular students. You can download the paper from the journal Science Education, volume 91, number 5, pages 805-821. [Access through university library.]

Triangulation of Investigators

Using this approach, two or more researchers are employed to carry out the same tasks, be it observation, interviewing or analysis. The rationale is to reduce bias and increase reliability – that is, the extent to which a research tool produces the same results irrespective of when, where or by whom it is conducted. A good example can be found in a paper by Lietz et al (2006) which you can download from the journal Qualitative Social Work.

These authors wanted to explore ways of establishing rigour and ‘trustworthiness’ (the closeness of fit between respondents’ meanings and the study’s findings) in qualitative research.  In this autoethnographic study of spiritual identity, they used a number of checks including triangulation by observer, in this case meaning inviting a researcher outwith the project team to conduct a secondary data analyses. On p. 451 – 453, the authors discuss the similarities and differences which emerged in the two analyses, leading to ‘further reflexivity’ regarding the impact of individual researchers’ identities on the analytic process. 

Triangulation Of Data

Denzin notes that triangulating data sources allows the researcher to use the same method to maximum theoretical advantage. Choosing dissimilar settings to study the same phenomenon (he gives the examples of the various meanings ascribed to death within different parts of a hospital), allows the researcher to identify which explanations, or aspects of them, hold true across the board and which are context-specific. 

There are three sub-types in this category - time, space and person. Thus, data can be collected about different people doing the same activity; it can be collected at different times of the day or night and it can be collected in different places. Denzin also identifies three levels at which data can be analysed – the aggregate, where data were collected from separate, unrelated individuals, the interactive, where interaction between people or groups is the focus of analysis and the collectivity, where the unit to be observed is a group, community or society. The most complex form of data triangulation combines some or all of the sub-types with some or all of the levels of analyses.  

Theoretical Triangulation

If methodological triangulation is the most satisfying, then ‘pitting alternative theories against the same body of data’, as Denzin describes theoretical triangulation, is the most difficult. Denzin admits that few studies achieve it and the example he offers is a hypothetical one.  He argues that triangulating theory will avoid the risk of researchers reaching atheoretical conclusions, selecting only those data which suit their pet views or developing small scale theory which has little relevance beyond the immediate situation. To triangulate theory, the researcher should

  • Draw up a list of all propositions which might explain, or have relevance to,  the research problem to be investigated
  • Identify the different ways each proposition might be interpreted
  • Carry out the research to see which propositions hold water
  • Discard those which proved untenable
  • Carry out further research to identify the most likely interpretations of those propositions still in the ring
  • Review the propositions which passed and failed the empirical test, along with their parent theories
  • Arrive at a new theoretical understanding of the problem. Depending on the outcome of empirical work, this might include insights from a number of apparently conflicting theories.

This form of triangulation is another version of the dialectical position we looked at in Topic 1, illustrated by the excerpt from Jennifer Mason’s paper. 

A Cautionary Note

The concept and practice of triangulation has been current in social and educational research for decades and, as seen above, has a specific meaning, aims and procedures. Over the years, however, the term has been used widely and loosely: many studies claiming to use triangulation bear little resemblance to these principles and strategies. Sandelowski (2003: 328) commented that ‘having too much meaning, the word triangulation has no meaning at all.’ It has also been proposed that the term ‘triangulation’ be replaced by ‘crystallisation’ on the grounds that (at least in the field of post-modernist ethnography) the crystal is a more apt metaphor than the triangle (Richardson and St Pierre (2005). The latter is a ‘rigid, fixed, two-dimensional object’ whereas crystals are

‘prisms that reflect externalities and refract within themselves, creating different colors, patterns, and arrays casting of in different directions. What we see depends on our angle of repose – not triangulation but crystallisation.’ (p. 963).

Therefore if you intend to use triangulation, it is important to be clear about which form you are using, why and how you will do it - and to write that up clearly in your research report or dissertation. 

Further Reading

Read the following article by Stephen Gorard and Chris Taylor: “What is ‘triangulation’?” Building Research Capacity, February 2004, Issue 7, pp 7-9

Task 2

Kenewell et al carried out an investigation into The Use of ICT to improve learning and attainment through interactive teaching (2007).

How do their research design and analysis demonstrate principles of triangulation?