Triangulating Qualitative and Quantitative Approaches for the Analysis of Textual Materials

Mixing Content and Discourse Analysis as a Methodological Alternative

In order to understand the specific use we have done of the software T-Lab (Lancia, 2012; see Table A1 in Appendix), it may be useful to start with a brief methodological consideration concerning the triangulation of different analysis tools, and, in details, on the possibility to triangulate discourse and content analysis.

In general terms, when we talk about a mixed-method approach or triangulation (Fielding & Fielding, 1986), we refer to the idea of melting different points of view and perspectives, by using different researchers, different research paradigms, or different data analysis tools.

By proposing the triangulation of methods, we intend to propose a third way that is rarely approached and estimated, beyond the qualitative and quantitative analysis (Bazeley, 2002). In addition, when we discuss about textual material, we have to say that the nature of data is qualitative per se. In other terms, we deal with words and linguistic representations and not with mere scores at a given test or questionnaire, nor with standardized responses on a Likert-type scale. Having this premise in mind, there are always, at least, two different ways of approaching such data. On one hand, the researcher can focus on the way by which discourses have been done, with the implicit belief that what does matter is the how something is said; in other terms, what makes the difference are metaphors and peculiar word representations able to vehiculate power (think about the theoretical assumptions of the critical discourse analysis). On the other hand, the same textual material can be approached making reference to the repetitions of singular words and of word associations, assuming that what counts is quantity. In other words, even with qualitative data set, like natural conversations’ transcripts or textual materials, is still possible to decide between a qualitative approach and a quantitative one. Actually, a third alternative does exist: that is to say, the decision to take advantage of both the qualitative and quantitative perspectives on texts, assuming that both the different forms of repetition and the peculiar way by which something has been said may play an important role. In the present article, we will discuss about this latter possibility, presenting a data set approached along with the two different perspectives, that one qualitative, anchored to discourse analysis, and that one quantitative, the content analysis.

The first proposal of triangulation has been, at least at our knowledge, that of Wilson, who states: “qualitative and quantitative approaches are complementary rather than competitive methods ( … and the) use of a particular method ( … ) rather must be based on the nature of the actual problem at hand” (1981, p. 58). Such a proposal will be refined by the so-called methodological appropriateness paradigm developed by Patton (1990), according to which the choice for a specific analysis tool or technique has to respond not to the researcher’s limits but rather to the reality; in other words, it is the object of analysis to suggest the right method of analysis.

Utilizing T-Lab for the Study of Qualitative Data in Applied Social Sciences: A Brief Review of Triangulated Studies

The interest toward automatic textual analysis is on increase, ¹ as it is the debate about it. ²

On one hand, it is doubtless that the process of analysis can gain in transparency and replicability by adopting ad hoc softwares (Hwang, 2008) as well as in the data storing saving a lot of time but, on the other one, a “software is, even now, quite limited in the kinds of support it offers for analysis, and there is no prospect that it will ever excuse the need for researchers to think” (Fielding, 2002, p. 162).

We do agree with Hwang (2008, p. 524) when he states “that QDA software can be used at any point along the qualitative research spectrum, whether it is close to the theory-building or grounded-theory end or close to the theory testing or hypothesis-testing end. For practical benefit, using QDA software will save time and make your work easier to manage, especially for large sets of data and teamwork,” reaching the conclusion that the software can be seen as a research assistant.

Unfortunately, if we move to discuss about the possibility to adopt mix method by triangulating automatic content analysis and manual discourse analysis, we have to state that there is a paucity of studies realized by mixing different analysis tools applied to the same textual data set (see Manuti, Cortini, & Mininni, 2006, for an interesting mixing of metaphor analysis and diatextual analysis). The reason why for such a paucity has to be searched in the methodological conflict between qualitative and quantitative approaches, even when applied to qualitative data set, as we have already stressed elsewhere, talking about an ideological battle (Verrocchio, Cortini, & Marchetti, 2012).

If we look to websites content, for example, few studies have triangulated content analysis with discourse analysis; an interesting exception is given by the article of Annese and Traetta (2012), which triangulates in creative way content analysis with social network analysis. For what concerns articles triangulating different techniques and using T-Lab, we have to cite the article by Benevene and Cortini (2010), which suggests in explicit terms that the object of analysis should shed light on the methodological tools to be adopted, showing how fruitful could be mixing nomothetic and idiographic approaches to the same research object.

An additional article that used T-Lab in conjunction with additional textual analysis tools is that one by Di Martino and Zan (2010), who, referring to Lieblich, Tuval-Mashiach, and Zilber (1998), present an analysis of students’ narratives focused on math learning, done along two different factors: holistic—categorical and content form, supported by different T-Lab’s applications.

In order to show how T-Lab software may be employed in the triangulation of methods, we propose a brief research report: in other words, we don’t point out every software aspect or possible analysis, but simply illustrate how to use certain functions of T-lab following the perspective of mixed approaches of analysis.

Materials and Method

This study aims to explore which are the most recurrent themes, through a quali-quantitative analysis, of the site www.strandeast.com content (the official website of the Strand East) on September 30, 2012.

Strand East is an urban regeneration project of a East London neighborhood realized by Inter Ikea Group; the plan provides the transformation of an East London disused industrial area (named Dane’s Yard) in a waterside quarter based on the ecological philosophy of the Swedish Furniture Company.

The theoretical premise we start from is that one of narrative analysis applied to organizational contexts (Manuti & Mininni, 2013), according to which organizations live by discourses, so that, researchers should be ready to disentangle the network of collective narrations and discourses which is shaped through and by the shared practices of accounting (Cortini, 2009). We will analyze, in our case, the self-presentation of the project Strand East done on the official website, by the aid of the software T-Lab, providing the following analysis: word occurrences and co-occurrences, correspondence analysis, Markovian chains analysis.

Results

Before going into deep details of analysis, it is important to remark that we have prepared our text for analysis doing what is known under the name of lemmatization, which gives a reorganization of the T-LAB database, creating different tables used to analyze data; in particular the idea is that of cluster together words that have the same meaning root, like, for example, “family” and “familial.” Such an operation, of course, has been done only for the words (lemmas or categories) considered interesting for the subsequent analyses, like “city,” “sustainability,” “transport,” and so on.

Results of the Analysis of Word Occurrence and Co-Occurrence

The first thing T-Lab allows to do with textual material is to analyze word occurrences and co-occurrences.

The output of the software (Figure 1) shows in the middle the most cited word, and all around the words that co-occur the most with it, according to an association index: the Cosine coefficient; in graphical terms, the more two words co—occur, the more they are closed in the dimensional space.

OPEN IN VIEWER

Figure 1.

Word association of the most cited word: Strand East.

It is always possible to “dialog” with the software and asking to put in the middle a specific word of interest for the user in order to have a graphical representation of its associations; in such a sense, T-Lab can assist the user following both an automatic analysis path and a customized one. For what concerns our study, it is remarkable that “Strand East” is the most cited word; in other words, the website is highly autoreferential. By clicking on the words associated with the central one, it is possible to get the phrase where the two words co-occur (Figure 2) and this cue is particularly useful in terms of mix method, because just with a click you get the original textual material that can be analyzed by discourse analysis.

OPEN IN VIEWER

Figure 2.

Analysis of word associations: Strand East and home.

Results of Correspondence Analysis

Correspondence analysis allows the extraction of new factors that summarize in an organized way the significant information of the data table cells; furthermore, this analysis technique allows the creation of graphs showing—in one or more spaces—the points that detect the objects in rows and columns, that is—in our case—the linguistic entities with the respective source features.

For what concerns the correspondence analysis, we can extract two main factors (Figure 3), namely the space, which goes from the private dimension of “building” or “bedroom zone” to the public sphere of “olimpic park” or “tower.” The second factor refers to the idea of development, which goes from the dimension of economic “investment” to the idea of “planning” and “project.”

OPEN IN VIEWER

Figure 3.

Correspondence analysis output.

Results of Sequence Analysis

This first result pushed us to investigate in finer details the word associations of the lemma “Strand East”; for this reason, we run a sequence analysis (Figures 4 and 5).

OPEN IN VIEWER

Figure 4.

Markovian analysis of the sequences containing Strand East: predecessors.

OPEN IN VIEWER

Figure 5.

Markovian analysis of the sequences containing Strand East: successors.

Sequence analysis allows a Markovian analysis of the sequences concerning the lexical units in the network defined by the corpus. Sequences are syntagmatic relationships between the lexical units under analysis, each of them—for each occurrence within the corpus chain—has a predecessor and a successor, that are, respectively, the lexical unit that comes before it and the lexical unit that comes after it.

Beginning from a matrix in which all the predecessors and all the successors of each lexical unit are recorded, T-LAB calculates the transition probabilities (Markov chains) between the lexical units analyzed, providing also a table that shows the sorted list of predecessors (the first) and successors (the second) of the lexical unit under analysis, which is, in our case “Strand East.” We decided to explore predecessors and successors because they allow a better interpretation of the way by which a peculiar concept is represented, stressing, potentially, causes and outcomes (as they are perceived by speakers) of phenomena under investigation.

The results of the sequence analysis show the use of a series of interesting predecessors, which could be summarized into two separated clusters: the first one refers to the geographical area, with words like “site,” “London,” or “Stratfort.” It seems, in such a way, that the first aim is to insert Strand East in a specific area, something well known in house marketing, where the first need for potential clients is to understand where the building is sited. The second cluster refers to the idea of planning and vision, provided by words as “planning” and “vision,” framing Strant East as a visionary project.

For what concerns the successors, it is enlightening the probability of “home” and “house,” as well as that ones of “investment” and “opportunity”; the first two refer to the idea of having a residential area, not industrialized but near the city, while the latter ones remark, taking advantage of the recency effect (being successors), the opportunity to be taken.

Finally, it is remarkable that a cluster of words do appear both in the predecessor and in the successor lists, and these are all lemmas related to business, like “commercial,” “business,” “plan,” or “development”; such a circularity seems to stress that Strand East took origin as also a business project but, most important, it will bring and support business. In marketing terms, Strand East has presented itself as a social responsibility project, in favor of families, but it could be seen as a mere business affair. Perhaps it is not a case that, also in the official website, Strand East declares its association with business but it is even more interesting that business growth, development, and planning are primarily in the successor list, in order to guarantee the future clients economic development and welfare.

Limitations

If we look at the present article as a methodological article whose first aim is to present the potentialities of T-Lab, it is obvious that the first limits lie in the choice for specific analysis T-Lab provides, silencing all the other ones. In particular, among different alternatives of researching, we have not explored the linguistic specificity analysis, very useful when comparing two or more speakers or categories of speakers (think about comparative design in search for gender differences or aging differences or cultural differences).

On the other hand, if we look at the present article as a brief report on the self-presentation of the Strand East, it seems urgent to take into consideration the specular point of view, that is to say that one of IKEA consumers. It seems to us mandatory to enlarge the research design as to comprehend the feeling of real Ikea consumers in face of Strand East Project and considering the way by which they are influenced by such a project. In addition, in more applicative terms, it would be important to understand better what happens in consumers’ minds when companies adopt a large Corporate Social Responsibility project like Strand East is.