I use NVivo computer-aided qualitative data analysis software as part of my literature review process.This post describes some of the things I do as part of that process.* You can see an accompanying SlideShare here.
Background: My research studies PLT practitioners’ engagement with scholarship of teaching and learning in PLT. This study involves individual and extra-individual dimensions of practice. I am interested in how social structures (such as “law”) are inscribed into practices, and whether practices can affect or change those structures. As part of my research, I have collated about 2000 items of literature, which I need to organise effectively for my doctoral research and for the future.
I use Endnote, and I am pretty happy with it. I’ve observed, however, that different citation managers suit different styles of working. I will be talking about Endnote in this post, but the process I describe is probably similar to other citation managers.
I collect all my references in Endnote, along with copies of articles etc in PDF format. I write, or dictate, my own notes into the “research notes” field in an item’s Endnote record. I use Endnote’s “smart groups” feature to group references together by search terms that I specify. For example, I might want to group all references that mention “bricolage” in conjunction with Bourdieu. I can specify that Endnote search all records and attached PDFs for those terms and keep a record of the search results in a smart group labelled “Bourdieu and Bricolage”.
Smart groups help to organise the reference library around themes to which you frequently return during some stage of your research. They also provided a convenient way to select a list of references for export into NVivo for further analysis.
To export the selected references, I go to File, Export, and then choose to export the references in XML format, and save the file on my desktop for the time being.
Organising Literature in NVivo
I import the XML file into NVivo using the External Sources / Other Sources / Endnote dialogue. You can choose to arrange the references by Author or Title. NVivo will import the PDF files as internal sources, and all the notes attached to the Endnote records, such as abstracts and your research notes will be saved as linked Memos. You can add to these memos as you work on your literature review.
By default, Endnote will classify these sources as “Reference” and there will be a number of default attributes, e.g. author name, date of publication, etc. You can edit the classification sheet if you want to add extra attributes, or perhaps simplify it.
It is worthwhile paying attention to classifications and attributes for any sources, because you can use some powerful search and analysis tools on these attributes later.
Each of the references you import can be “coded” as “case nodes”. You can think of a “node” as a bucket for collecting things based around a concept. In the context of a literature review, you can code a reference, such as Bourdieu’s ‘Force of Law’ article, as a “case”. One benefit of doing this will be the ability to cross-reference “theme nodes” with all or selected “case nodes” later. For example, later I might code parts of references to the theme node “resistance to change”, I can then compare the text coded to the theme (or concept) “resistance to change”, to all the authors coded as cases.
Coding – Explicit and Emergent Approaches
“Coding” here, in its simplest form, involves selecting/highlighting text in the reference source, and linking it to a concept or theme. NVivo will keep a database record of all the bits of text (or pictures or video) that are “coded” to a concept. Any one portion of text can be coded to multiple concepts, so you’re not shut out from revisiting a particular item from different points of view.
The decision to categorise a concept or theme might be described as “explicit”, or as part of an “emergent” approach. Explicit and emergent approaches are identified with the methodology described as “grounded theory”.
You might already have a clear theoretical framework that you are using, with explicit themes or concepts. You can predefine the nodes (remember the “buckets”?), according to your explicit concepts. For example, you might have predefined concepts of “motivation”, “capability”, “symbolic support”, “allocation of resources”.
On the other hand, you might choose to see what concepts “emerge” from the data, and define these as you work through the sources.
For literature reviews, I personally tend to explicit approaches at first, but bearing in mind the idea of “theoretical sensitivity” that seeks to accommodate both explicit and emergent approaches.
Coding from Text Search Queries
If you have some explicit concepts for your literature review, it could be worth starting with a text query, then “code on” from the results using in-vivo techniques.
For example, I could search for the word “bricolage” in all the references, and set the text search dialogue to code and save all the results to the results node “bricolage text search”. The results will be saved in the Query/Results folder.
When the search is complete, a results screen will open showing me all the “hits”. I can see these highlighted on the PDF source, and read around the hit to understand the context in which the word is used. If I identify further concepts or themes, I can select that text and code it to a new theme or concept node. For example, some text might show how “bricolage” is “emancipatory”, so I could create a new theme node called “emancipatory” and code the relevant text to that node. In doing so, I am coding emergent concepts as I go along “in-vivo”.
Let’s say that I’ve collected quite a few theme nodes, but now I want to compare how the authors relate to one or more of the themes. Using the framework matrices tool, I can arrange all the cases (authors) in table rows (remember I coded all the references as cases earlier?). I can arrange one or more themes or concepts in table columns. NVivo will then search the selected sources, and using the “autosummarize” feature, populate the cells of the table with the text code to the specific author and theme node.
So, for example, I can compare the text coded to “emancipatory”, author by author.
You can format and edit the text in the framework matrix cells. You can also continue to “code-on” the text in the framework matrix cells.
You can sort the cases by attributes too. Say for example you identify certain authors from different “schools” of thinking, or affiliated with specific institutions. You can record these attributes, and then sort the framework matrix results by those attributes.
Matrix Coding Queries
Matrix coding queries allow you to compare sources and theme nodes by attributes. So, for example, I could compare all sources published before and after a certain date against a specific theme or concept. Or I could compare theme nodes against other theme nodes, for example I could compare sources coded to theme “emancipatory” against the theme, “practical knowledge”, to see if those terms appear more or less frequently in some sources than others. Matrix Coding Query results can be charted, and these can generate further lines of inquiry. For example, you might notice the intersection between two concepts “pile up” around a specific author or group of authors, or are ignored by other authors.
The charts that NVivo produces are “live”. This means you can double-click on a point in the chart and it will open the relevant references connected to that result.
I have briefly described some of the NVivo tools I use when undertaking literature review. These tools are quite powerful, and if you consider that you might apply them to dozens of sources at a time, you might apprehend how you can expedite parts of your literature review. That said, it is just a tool and no substitute for your own critical and intellectual engagement with the source material.
You might also think parts of this literature review process is very useful for lawyers in researching case law, statutes, trial preparation and proof-making.
In a later post I will describe how I export the results and the notes recorded in memos in a way that allows me to easily incorporate them in writing up the research, by using a publishing tool called “Scrivener”.
*This post is based on a presentation I gave at Deakin University’s Faculty of Arts and Education “Methodology on Fridays” session on 7 March 2014.