18.4 Thematic analysis

What are you trying to accomplish with thematic analysis?

As its name suggests, with thematic analysis we are attempting to identify themes or common ideas across our data. Themes can help us to:

• Determine shared meaning or significance of an event
• Povide a more complete understanding of concept or idea by exposing different dimensions of the topic
• Explore a range of values, beliefs or perceptions on a given topic

Themes help us to identify common ways that people are making sense of their world. Let’s say that you are studying empowerment of older adults in assisted living facilities by interviewing residents in a number of these facilities. As you review your transcripts, you note that a number of participants are talking about the importance of maintaining connection to previous aspects of their life (e.g. their mosque, their Veterans of Foreign Wars (VFW) Post, their Queer book club) and having input into how the facility is run (e.g. representative on the board, community town hall meetings). You might note that these are two emerging themes in your data. After you have deconstructed your data, you will likely end up with a handful (likely three or four) central ideas or take-aways that become the themes or major findings of your research.

Variations in approaches to thematic analysis

There are a variety of ways to approach qualitative data analysis, but even within the broad approach of thematic analysis, there is variation. Some thematic analysis takes on an inductive analysis approach. In this case, we would first deconstruct our data into small segments representing distinct ideas (this is explained further in the section below on coding data). We then go on to see which of these pieces seem to group together around common ideas.

In direct contrast, you might take a deductive analysis approach (like we discussed in Chapter 8), in which you start with some idea about what grouping might look like and we see how well our data fits into those pre-identified groupings. These initial deductive groupings (we call these a priori categories) often come from an existing theory related to the topic we are studying. You may also elect to use a combination of deductive and inductive strategies, especially if you find that much of your data is not fitting into deductive categories and you decide to let new categories inductively emerge.

A couple things to note here. If you are using a deductive approach, be clear in specifying where your a priori categories came from. For instance, perhaps you are interested in studying the conceptualization of social work in other cultures. You begin your analysis with prior research conducted by Tracie Mafile’o (2004) that identified the concepts of fekau’aki (connecting) and fakatokilalo (humility) as being central to Tongan social work practice.^[1] You decide to use these two concepts as part of your initial deductive framework, because you are interested in studying a population that shares much in common with the Tongan people. When using an inductive approach, you need to plan to use memoing and reflexive journaling to document where the new categories or themes are coming from.

Coding data

Coding is the process of breaking down your data into smaller meaningful units. Just like any story is made up by the bringing together of many smaller ideas, you need to uncover and label these smaller ideas within each piece of your data. After you have reviewed each piece of data you will go back and assign labels to words, phrases, or pieces of data that represent separate ideas that can stand on their own. Identifying and labeling codes can be tricky. When attempting to locate units of data to code, look for pieces of data that seem to represent an idea in-and-of-itself; a unique thought that stands alone. For additional information about coding, check out this brief video from Duke’s Social Science Research Institute on this topic. It offers a nice concise overview of coding and also ties into our previous discussion of memoing to help encourage rigor in your analysis process.

As suggested in the video^[2], when you identify segments of data and are considering what to label them ask yourself:

• How does this relate to/help to answer my research question?
• How does this connect with what we know from the existing literature?
• How does this fit (or contrast) with the rest of my data?

You might do the work of coding in the margins if you are working with hard copies, or you might do this through the use of comments or through copying and pasting if you are working with digital materials (like pasting them into an excel sheet, as in the example below). If you are using a CAQDAS, there will be a function(s) built into the software to accomplish this.

Regardless of which strategy you use, the central task of thematic analysis is to have a way to label discrete segments of your data with a short phrase that reflects what it stands for. As you come across segments that seem to mean the same thing, you will want to use the same code. Make sure to select the words to represent your codes wisely, so that they are clear and memorable. When you are finished, you will likely have hundreds (if not thousands!) of different codes – again, a story is made up of many different ideas and you are bringing together many different stories! A cautionary note, if you are physically manipulating your data in some way, for example copying and pasting, which I frequently do, you need to have a way to trace each code or little segment back to its original home (the artifact that it came from).

When I’m working with interview data, I will assign each interview transcript a code and use continuous line numbering. That way I can label each segment of data or code with a corresponding transcript code and line number so I can find where it came from in case I need to refer back to the original.

The following is an excerpt from a portion of an autobiographical memoir (Wolf, 2010)^[3]. Continuous numbers have been added to the transcript to identify line numbers (Figure 18.4). A few preliminary codes have been identified from this data and entered into a data matrix (below) with information to trace back to the raw data (transcript) (Figure 19.1).