Glaser and Strauss developed the Grounded Theory (GT) approach in the 1960s and thus before the event of CAQDAS. Strauss and Corbin developed their version of it in the 1980 and 1990. Anselm Strauss was introduced to earlier versions of ATLAS.ti. In the preface of the ATLAS.ti version 4 manual (Muhr 1997, S. 1) he wrote:
Now the program has been further elaborated and improved. It should be of considerable aid in providing both greater efficiency and more elaboration for social scientists who master its intricacies. I myself am no expert computer-based researcher, and at my age am not likely to become one.
Juliet Corbin also sees the benefits of computers to support a GT analysis. She writes: The computer has the ability
to augment the human mind by doing a lot of the detailed and tedious work Involved in many endeavors, thus freeing up the user to be creative and thoughtful. And this is what computer programs do for qualitative analysis” (Corbin & Strauss 2015: 203).
Even though the authors acknowledge the usefulness of software for the analysis process, the various steps and procedures for a GT analysis are only described for manual ways of going about analysis with a few pointers how this might be done using CAQDAS.
For young researchers being digital natives it is hard to image not to be utilizing computer technology, be it for data collection, transcription or analysis and reporting. However, learning the various tools and features in a software, does not automatically teach them which tool is the best fit for which step in the analysis process given a specific methodological framework like Grounded Theory. In this article, I therefore would like to offer a translation of the process of open coding using ATLAS.ti. It is more than just clicking on the “open coding” button in ATLAS.ti.
The first question one needs to ask is what coding means when using ATLAS.ti and what ‘open coding’ means from a Grounded Theory perspective. The use of the words “code” and “coding” in qualitative data analysis is a likely result of the widespread adoption of the Grounded Theory approach. It has also been embraced by almost all CAQDAS developers and those writing about software application for qualitative data analysis. But what does “coding” mean in CAQDAS? It simply refers to the process of attaching a label to a data segment. Computer scientists call this “tagging”– and for the sake of this article, it is the much better term to use in order not to confuse it with the much more complex process of open coding in Grounded Theory.
I need to point out that GT coding cannot be generalized across the various GT approaches that exist today. Strauss (1998) gives an example of open coding. The result is not a list of tags, but a memo of several pages in length that holds words that stand for categories, properties, or dimensions. Charmaz’ initial coding process resembles much more what CAQDAS users normally do when tagging data. For her the way of labeling makes the difference between grounded theory coding and general qualitative coding. Grounded theorists code for actions rather than topics. For instance, rather than “Friends’ support” or “Hospitalization”, a grounded theorist would use initial codes like “Receiving friends’ help in seeking care / Requesting regimen re-evaluation”, or “Gaining medical access / being admitted to hospital” (Charmaz, 2014:122-123).
Both Strauss (1987/1998) and Corbin & Strauss (2015) mentioned that it is a possibility to write the concepts that they usually write on index cards, as labels in the margin of a document. Strauss assumed that this was probably common practice but pointed out these “codes” would then be less detailed and more difficult to sort (Strauss 1998: 114). Looking at the examples Strauss provides, the notes on his index cards often hold quotes from the data, references to other related data segments, analytic reflections, pointers for theoretical sampling, and references to the coding paradigm. Further, he wrote down on the index cards whether within a coded segment there were references to a condition, an interaction, or a consequence are contained, i.e. axial coding. This also applies to the examples given by Corbin & Strauss (2015). Thus, when using the word coding in the context of CAQDAS is not what Strauss and Corbin mean when they talk about coding in the Grounded Theory sense. GT coding is more than just attaching a label to a data segment.
To demonstrate the process of open coding, I use the data of the sample project that Juliet Corbin uses in the 2012 and 2015 editions of her book Basics of Qualitative Data Analysis. The data consist of an open interview carried out by Anselm Strauss in 1987 with a Vietnam veteran who worked in the medical corps; two others, one semi-structured with a US Marine who served in the Vietnam War, and an email conversation with a Panama, Saudi and Bosnia veteran, also US Marines, were conducted by Juliet Corbin in 2006. Initially, there are no detailed or fixed research questions. There is only a general broad interest in the experience of a soldier in the war.
During the first phase, I read the data and recorded the ideas and thoughts that occurred to me using quotation comments. In addition, I tagged the data. My intention was to pre-structure the material to take advantage of the computer early, in that it allows me quick access to the material for further analysis via tags. The initial list of tags is shown in Figure 1. The tags are not yet sorted or colored. At this stage of analysis, I only created a list of terms, for the moment ignoring any possibility to sort tags for hierarchical structure. Please also note, this first step of tagging is not the equivalent of open coding. Going through and tagging the data can be an initial step in order to make it easier to move through the data and find things, but in order to label it open coding according to the Strauss /Corbin approach, an essential step is missing – namely writing. For this, I made use of the quotation comments.
ATLAS.ti offers a level of analysis below tagging that facilitates the process of GT open coding. A coded segment consists of what ATLAS.ti calls a quotation and a code linked to it. Thus, a quotation is an entity of its own and because of this several actions are possible: A quotation can be named, commented, it can be linked to other quotations, and one or also multiple memos can be linked to it. Further in terms of functionality, the ATLAS.ti margin area is an essential feature especially when using an interpretive approach. It gives immediate feedback throughout the analysis how a segment is tagged, whether it is linked to other segments, whether comments or memos have been written, or whether it is part of a network (see Figure 2).
Another ATLAS.ti feature that supports data level work is the ability to create named links between quotations called ‘hyperlinks’ When creating a hyperlink in ATLAS.ti, i.e. linking two or more quotations to each other, the researcher specifies how the data segments relate to each other: A explains B, A discusses B, A contradicts B, A is a supporting statement of B, etc. Figure 3 below shows the ATLAS.ti Hyperlink Manager and five links that were created in the process of tagging the first interview. In the text itself, you can jump back and forth between hyperlinked quotations via double-click on a hyperlink, just as if you clicked a link on the Web. The link can be between two quotations only, or continued to form a chain or star links. This means it is not necessary to write down references on index cards as in manual analysis to search for them when needed—they are available instantaneously.
After tagging the first interview, I began with reviewing the code list adding some structure to it. Next, I tagged interview two and three continuing the interpretive process in writing quotation and code comments and linking quotations to each other as relations emerged. Figure 4 shows the changes to the tag list that resulted from this. The first two lists expand and clarify the open coding process as defined by Corbin & Strauss: open coding is the process of “breaking apart data and delineating concepts to stand for blocks of raw data. At the same time, one is qualifying those concepts in terms of their properties and dimensions.” (Corbin & Strauss 2008: 195). Thus, the initial tag list turns into a category system over time where tags receive methodological significance.
The software entity ‘code’ needs to serve two purposes for GT: an initial tag can be developed into a category, a property, or a dimension and thus takes on methodological meaning. Some ‘codes’ however may never be methodological; they simply serve organizational purposes to support data retrieval. Further, the code system needs to be developed and structured in a way to be able to use the advanced software tools later in the analysis.
For the next level of abstraction, I used the code comment area. The purpose was to summarize the content of the tagged quotations as a step towards moving from a descriptive to the conceptual level. In addition, I wrote two memos. One was about background information on traumatic experience, as this is well researched. And the second one on Roger’s congruence theory, as I was reminded of it while reading the data. If I had not prepared this data set as an exemplary case study, I would have written other memos for instance to outline the project idea, to write down my own thoughts and position regarding the topic at hand, etc. The availability of the quotation level and the comment function plus having memos as containers for writing, allowed me to distinguish between various methodological activities. From a GT perspective, most information written in ATLAS.ti comments would be considered memos. But if you were to use ATLAS.ti memos for all of your writing, because this is what it is called in GT, you would not utilize ATLAS.ti to its fullest potential. My experience is that users often stop writing memos, if they use memos for all levels of their analysis because they generate too many memos and then get or feel lost.
If you are interested in how I continued the analysis of this sample data, a full paper (Friese, 2016) can be downloaded here: https://www.mmg.mpg.de/61762/wp-16-07
The ATLAS.ti quotation level can also be utilized translating the Charmaz approach to Grounded Theory. Instead of using tags for initial coding, I would recommend creating quotations at this first stage of analysis and rename the default name. The quotation name thus becomes the ‘code’. The actual software entity ‘code’ comes into play during the process of focused coding. This way it is possible to preserve all thoughts that went into initial coding and represent focused coding within a different software tool (Figure 5). Analysts may also be tempted to leave the initial codes in the code list rather than to merge them, ending up with a very long list of codes resulting from different levels of analysis.
Glaser (1978) and also Charmaz (2014) advise to code with gerunds because it helps researchers to detect processes and to stick to the data. In a computer-assisted analysis, given the multiple functions of tags that also need to serve as codes in a methodological sense, there is no straight-forward general translation for it. When starting to sort tags in ATLAS.ti, it is best to work with prefixes. In addition, the alphabetic order needs to be observed. This makes it difficult to start all code labels with a gerund. A gerund might occur as part of the label, or can be used when describing the code in the code comment.
In this article, I have shown how to translate the Strauss /Corbin open coding process in ATLAS.ti and I sketched out which functions can be used for initial and focused coding according to Charmaz. The take-home message from this article is that you first need to understand the meaning and purpose of a given methodological task. The second step is to apply the proper software functionality to realize the task. This often involves a process of translation. A code or a memo in a methodological sense is not necessarily what is called a code or a memo by a software tool. If you are interested in reading more about how to go from methods to implementation, in their upcoming 5-level QDA books, Woolf and Silver (2017) offer their interpretation of the process of translation. See: http://www.fivelevelqda.com.