Theme and Category Development in ATLAS.ti
Author: Susanne Friese
In principle there are three ways to begin coding. You either start with a list of codes or you develop your code list from scratch – or use a mixture of the two. Initial ideas for coding might directly emerge from the data you read, or can be derived from research questions, theoretical models, literature or from an interview guideline. When you start coding, the how descriptive or abstract a code is does not play a role; codes may be descriptive or already conceptual. The important point is to mark those things that you find interesting and to name them.
Reading further, you will very likely notice a few things which are similar to some issues you have noticed before. These may either fit under an already existing label, or you may need to rename a label to subsume the already coded and the new data segment. Even if the term is not yet the perfect code label, it does not matter. You can continue to collect more similar data segments and later, when you review them, it will be easier to think of better and more fitting code labels. At first you will generate lots of new codes; in time you will reuse more and more of those codes that you already have. If this is the case, you have reached your first saturation point. In technical terms, you will drag and drop existing codes from the Code Manager or Navigator onto the data segments. As soon as you reach this point, it is time to review your coding system.
Based on the codes shown in Figure 1, we would like to demonstrate how an initial list of codes can be sorted and structured. The code list is based on the children and happiness sample project (See LINK). The version of the project we are using here contains 46 codes related to positive and negative effects of being a parent. These code labels are partly directly taken from the words the respondents used and thus are very close to the data. The frequency of each code word is very low.
What often happens at this stage is that analysts use code families to collect descriptive code labels without further conceptualizing. This is not a good idea. It leads to endlessly long code lists that are of no use in querying the data in the next phase of the analysis. Below we will make use of code families (code groups) as title helpers only.
If we look through the list of codes in the Code Manager, it can quickly be noticed that there are positive and negative effects of parenting. Therefore, all positive effects are grouped into a code family called ‘positive effects’. No we can apply a global filter with a right click on the code family (ATLAS.ti 7 Windows) to focus on the reduced list of 22 codes. In the Mac version, just click on the code group to filter the list of codes.
Looking at the reduced list of codes, it is now easy to see that a number of codes refer to personal growth. Personal growth already exists as a code. So we can merge those codes that are similar into this code. For this, do as follows:
- Right click on the code ‘personal growth’ and select Merge Codes. From the list that pops up select all codes that relate to personal growth. This reduces the list to 14 codes. The personal growth code now contains 8 quotations. In the Mac version drag the codes to ‘personal growth’ and select the merge option. You do this for one code at a time.
- Look at the remaining codes. Other codes you can merge are those related to fulfilment, richer life, positive emotions and improved relationships (parental and marital). If the more conceptual name does not yet exist, rename one of the codes with the conceptual term and then begin the merge process. This reduces the filtered list to 5 codes.
As these code labels would still appear all over the place in the alphabetic code list, we need to add a prefix to all of these labels, so that they are united under a common heading. I suggest to use the prefix ‘effects pos’ and for the second list ‘effects neg’, so that the effects of parenting codes are sorted underneath each other later in the complete list of codes. The last thing that is missing is the main category label. As we built the category from the bottom up, it does not yet exist.
- Create a new code with the name EFFECTS POS. Highlight all codes and give them a color.
- Take out the global filter (ATLAS.ti 7 Windows): right click on the code family, select Remove global filter. Click on the button Show all Codes.
Voilà, there is your new category with sub codes:
The goal in developing subcategories is to achieve a good description of heterogeneity and variance in the data material. We will again use an example based on the children and happiness project (download sample projects from here). This time, the data segments have been “lumped” under only a few codes. No attempt has yet been made to separate the various aspects.
When you begin to code, you may not know which aspects are salient and provide good reasons for forming a group of their own, so it is easier to collect the various aspects under a main theme first. But before there is too much coding, you need to begin to sub-divide it. I will explain the procedure based on the code “children are hard work but…”.
- Double click on the code “children are hard work but…” and read through the quotations. In the Mac version, select the code in the Code Manager and read the quotation in the pane below the code list. If you prefer reading the quotations in an editor, select the code and then the option Output–>Quotations for Selected Code(s). Your task is to develop subcategories based on the major theme.
After reading through a few quotations, you are likely to notice quotations that refer to the same aspect and others that describe another facet of the main theme. When you are at this point, start collecting these aspects. One option is to write them down on a piece of paper and run a tally. Figure 6 shows my notes. Given sufficient enough screen space, another option is to type your ideas into a memo and use this memo later on to create the list of sub codes (Windows version, see: ‘importing codes’ under Help).
Looking at the extracted terms helps to conceptualize them further. ‘Greatest joy in life’ might be subsumed under ‘richer life’. Would not trade it for anything, worthwhile and rewarding could be summarized under ‘worthwhile trade-off’; life is meaningful, I feel fulfilled can be integrated into fulfilment, and so on. The aim is to look at the bandwidths of issues mentioned and to come up with a label that best describes the most similar ones.
After you have decided which subcategories you want to use, you need to enter them into ATLAS.ti and then recode the data. In order for the sub-codes to show up under the main code, we need to add prefixes again and rename the main category name, as follows:
HARD WORK BUT…
Hard work but: fulfilment
Hard work but: richer life
Hard work but: worthwhile trade-off
Hard work but: positive emotions
In ATLAS.ti 7 Windows, click F5 to re-sort the list. Now it is a good time to colour the codes of this newly developed category. All quotations are still contained in the main category code. The next step is to distribute them to the sub codes (Figure 7).
- Double click on the main category code HARD WORK BUT… and select the first quotation.
- Read it and decide into which subcategory you want to move it. Select the appropriate sub code in the Code Manager and drag it onto the main code in the margin area, thus replacing it. Another option is to use the code list from the navigator pane.
There is no need to leave the main category code attached to the segment. It just clutters up the margin. It is easy to collect the quotations in one main category via a code family / code group later. For now the aim is to empty the main category code and fill the subcategories with content (see Figure 8). You may need to adjust the length of some quotations or create new ones. If you come across a quotation that you find difficult to sort into any of the existing subcategories, leave them in the main category.
In this article, I have shown the process of building categories and sub codes when approaching analysis inductively. The process is different if you chose a deductive approach. Knowing how an efficient code system looks like – efficient in terms of being able to utilize the more advanced analytic tools of the software, you can name known categories and sub codes already in the same way as for instance shown in Figure 9. If you need to complement the code list, you can apply the procedures as described above. The structure of the final code list will be the same, no matter whether you start inductively or deductively. As a reference example, please take a look at the Children & Happiness sample project and open the Code Manager.
Video tutorial on building an efficient coding system (11 minutes). Download from here.
Video tutorial on building categories (18 minutes). Download from here.
About the Author
Dr. Susanne Friese started working with computer software for qualitative data analysis in 1992. Her initial contact with CAQDAS tools was from 1992 to 1994, as she was employed at QualisResearch in the USA. In following years, she worked with the CAQDAS Project in England (1994 – 1996), where she taught classes on The Ethnograph and Nud*ist. Two additional software programs, MAXQDA and ATLAS.ti, followed shortly. Susanne has accompanied numerous projects around the world in a consulting capacity, authored didactic materials and is one of the principal contributors to the ATLAS.ti User’s Manual, sample projects and other documentations. In 2012 / 2014 her book “Qualitative Data Analysis with ATLAS.ti” was published with SAGE publications.