How To Do Axial Coding with Examples
This is part of our Practical Guide to Grounded Theory Essential | Start a Free Trial
What is Axial Coding?
Axial coding is the second coding step of grounded theory, where you begin to draw connections between ideas in your research. With grounded theory, you are looking to turn your qualitative data (such as transcripts from in-depth interviews or focused groups) into a new theoretical framework. This is done through the qualitative method of “coding”, where you label and organize your qualitative data with codes.
To learn more about grounded theory, check out our Practical Guide to Grounded Theory.
Axial coding in grounded theory is always preceded by another coding method, such as “open coding”. In contrast to open coding where you broke your data into discrete parts, in axial coding you begin to draw connections between codes. With Axial coding, you are identifying which codes from open coding are the most important and central to your theory, and refining and elevating them to the status of category.
To get a high-level view of open, axial, and selective coding check out our blog post on the topic.
Axial Coding in Qualitative Research
Axial Coding Definition
With axial coding in qualitative research, you read over the codes and their underlying data to find how they can be grouped and abstracted into categories. These categories could be created by abstracting out an existing code or developing new concepts that encompass several different codes. During axial coding, you will find some codes that closely resemble one another. These redundant codes can be merged and renamed to further organize your research.
After conducting axial coding you will have several categories that are supported by a cleaned-up set of supporting codes. These categories are the “axes” around which their supporting codes revolve.
Axial Coding Example
Axial Coding’s Coding Paradigm
With axial coding, you are looking to find “Categories” which are derived from the relationships between the codes developed in open coding. To help with this, Corbin and Strauss (1999) developed a Coding Paradigm that defined six subcategories. These subcategories, which are phenomenon, causal causation, strategies, consequences, context, and intervening condition, help ensure that you the researcher have fully explored the categories that you are developing. In the following example, we will be defining these subcategories, how to derive them from your open codes, and how to define your category based on them.
After we are done with the axial coding process, we will have a handful of categories, supported by the 6 subcategories, each of which have emerged from the codes we developed in open coding.
We don’t mean to be too prescriptive in this approach to grounded theory, as the whole point is not to impose a previous theoretical framework but rather develop your own from the data. Our human brains are naturally equipped to do pattern recognition, so many of these relationships will come naturally. However, Corbin and Strauss’s Coding Paradigm and the subcategories defined therein are a helpful tool to make sure you’ve fully explored the categories that you are developing in this axial coding phase.
Watch Webinar
See how axial coding works within the process of grounded theory
Before You Starting Axial Coding
To do axial coding, you must first have generated several initial codes, most commonly through open coding. Learn more about open coding in our Open, Axial, and Selective Guide.
Example of Axial Coding
Example: We will be referring to an example study where interviews were done with New Yorkers experiencing lockdown due to Covid-19. From this step, we developed several codes, including:
Breaking up the day
Experiencing a panic attack
Feeling disconnected
“Friends created online poker nights”
Introverted v Extroverted
Laughing leading to elation
Meeting for park hangouts
“Missing my family”
Moved back home
“Nervous about what’s going to happen”
Nervousness over in-person meetups
Only had the one panic attack
Past experiences with panic attacks
Rented an Airbnb
Researching how safe restaurants are
Scheduling virtual hangouts
Shortness of breath every day
Sleeping well on weekends
Therapy before Covid-19
“Trouble sleeping”
Waking up tired
“Wanted a hug”
Wishing for a larger apartment
“Worse than a long-distance relationship”
Zoom burnout
“Zoom dinner club”
Figure 1: A subset of the codes we developed in our Open Coding phase of analysis.
In the following steps, we will be modifying and grouping these codes into subcategories, from which we will derive one of our categories.
Subcategory 1: Phenomenon
Core to Corbin and Strauss’s Coding Paradigm is the concept of a phenomenon. The phenomenon subcategory is the central idea of a category and is the “what” that you will be exploring. One way to identify your phenomena is to read through the codes that you developed in open coding and try to identify common experiences.
Example: Reading through our codes that were developed during previous open coding rounds (See codes in Figure 1) from our Covid-19 Lockdown interviews we began to see a pattern amongst a set of codes that could hint towards a central phenomenon.
Open Coding Codes:
“Trouble sleeping”
Experiencing a panic attack
Waking up tired
“Nervous about what’s going to happen”
Re-reading the quotes from these codes, we observe that all of them point towards the research participants describing some form of anxiety. We create a new subcategory called “Phenomenon: Anxiety”, and nest the rest of them under it. With the final code setup looking like this:
Phenomenon: Anxiety
“Having trouble sleeping”
Experiencing a panic attack
“Nervous about what’s going to happen”
We now have a preliminary phenomenon that we can explore in the future steps.
Note: You may have noticed that our code changed slightly, it is a suggested practice to adjust and modify your codes from open coding as you work through them in the axial coding stage.
Subcategory 2: Causal Conditions
Now that you have uncovered a phenomenon in your codes, a good next step is to look for codes from your open coding step that could be the causal conditions of that phenomenon and group them under a sub-category that describes this causal condition. You are looking for codes from your open coding phase that answer the question “Why did this phenomenon happen?”.
Example: Having identified the phenomenon of “Anxiety”, we were now able to focus on the data that might describe “Why were people feeling anxiety”? In our list of codes from Figure 1 we saw codes like:
“Missing my family”
“Wanted a hug”
Feeling disconnected
“Worse than a long-distance relationship”
Reading over the data we found that the research participants were often directly connecting their loneliness to their own sense of anxiety. We defined a new subcategory “Cause: Lockdown Loneliness”, and grouped these codes:
Cause: Lockdown Loneliness
“Missing family”
Wanting physical connection
Feeling disconnected
Subcategory 3: Strategies
Thus far we have been talking about things that have happened to your research participant. They have experienced a phenomenon, and you have pinpointed one or many causes. Now it’s time to ask, what did the research participant do because of the phenomenon. What actions, or potential actions, did the research participant take. These actions are your research participants’ strategies. Create a subcategory based on these strategies and group together codes from the open coding round underneath it.
Example: Again reading over our codes and the underlying data in Figure 1, we start to see some potential strategies that our research participants used to address their “Loneliness induced Anxiety”. Codes like:
Researching how safe restaurants are
Meeting for park hangouts
Scheduling virtual hangouts
“Friends created online poker nights”
“Zoom dinner club”
Here we are seeing two potential strategies, as our research participants are trying to find ways to connect with other people both virtually and in person. With some organization we put all these under our subcategory: “Strategies: Covid-19 Conscience Connecting”
Strategies: Covid-19 Conscience Connecting
Attempting safe in-person meetups
Meeting at parks
Researching safety of indoor meeting
Setting up virtual hangouts
Games and clubs
Meeting over zoom
Theoretical Sampling - Collecting More Data
We interrupt this list of subcategories to talk about Theoretical Sampling. So you are working through these subcategories and suddenly realize you don’t have the data to support one of them. For example, perhaps you’ve identified a phenomenon but can’t find data that would support a causal condition. Or maybe you have identified some strategies, but you find that you have a research participant that is an outlier that took a different strategy from all your other participants. That is where Theoretical Sampling comes in.
Grounded theory is an iterative process, and analysis and data collection should be happening simultaneously. That is one of the major benefits of using the Coding Paradigm. It points out when you may have gaps in your theory. So when you notice that data supporting any one of these 6 categories is sparse, incomplete, or missing entirely, you should consider going back into the field and collecting more qualitative data. This act of targeted data collection based on gaps in your burgeoning new theory is called Theoretical Sampling.
Example: For the Covid example, we spoke with our research participants and had identified multiple participants who had set up Virtual Hangouts as a means to address their Anxiety. We however spoke with one participant that did not set up any Zoom calls but instead had been meeting friends in the park. Rather than writing this off as an outlier, it’s a sign that our strategies subcategory is incomplete. And we looked to speak with more people who used in-person hangout to address their anxiety. This will not only help us with our strategies subcategory but will help flesh out all of our subcategories.
Subcategory 4: Consequences
At this point, you have a phenomenon with one or many causes. In reaction to the phenomenon, your research participant developed and potentially acted upon a number of different strategies. The outcomes of those strategies are the consequences. Consequences can be both the actual outcomes of a strategy or the expected outcome of a strategy. So ask the question, what happened as a result of your research participants’ strategies? Create a subcategory that describes the consequence.
Example: Reading over your codes from open coding in Figure 1 and the corresponding quotes, you may identify these codes that describe the outcome from the strategies of “Attempting safe in-person meetups” and “Setting up virtual hangouts”:
Zoom burnout
Nervousness over in-person meetups
Laughing leading to elation
Breaking up the day.
You may notice while some of the consequences are beneficial, there also seemed to be some drawbacks as well. So we create a new subcategory called: “Consequences: Emotional Benefits and Drawbacks of Covid-19 Interactions”
Consequences: Emotional Benefits and Drawbacks of Covid-19 Interactions
Benefits
Laughing leading to elation
Breaking up the day
Drawbacks
Concern over contracting Covid-19
Screen time burnout
Subcategory 5: Context
Throughout all these steps you will probably already have started to identify codes that provide context to the phenomenon. The context is any number of details that describe the phenomenon or circumstances in which the strategies take place. This can be details like where the phenomenon happens or the intensity or frequency it occurs.
For example for the Covid-19 study we may have some of the following codes:
Wishing for a larger apartment
Moved back home
Rented an AirBNB
Shortness of breath every day
Sleeping well on weekends
Only had the one panic attack
These codes are providing context around the location where the participant is experiencing the anxiety and the frequency as well. In this case, we create two subcategories for context and group the codes under them.:
Context: Current Housing Setup
Living in apartment
Temporary living situation - moved home or Airbnb
Context: Anxiety Frequency and Intensity
Low grade but daily
In-Frequent but Intense
Subcategory 6: Intervening Conditions
Intervening conditions and context are similar, and some researchers use a simplified coding paradigm where they combine the two just into just “context”. Intervening Conditions are more general than context and describe attributes, usually about the participant, that may influence their strategies. This is often “background info” or previous experience. It can describe the participant, such as demographic information, or previous experiences before this specific phenomenon.
Example: As before we look to the codes from open coding to explore our intervening conditions. Intervening conditions will jump out often as quotes that describe the research participants' life before the phenomenon in question. We might have the following codes:
Past experiences with panic attacks
Therapy before Covid-19
Introverted v Extroverted
This we can rearrange them under these new subcategories:
Intervening Conditions: Mental Health Pre-Covid-19
Past panic attacks
Therapy pre Covid-19
Intervening Conditions: Personality
Introverted v Extroverted
Defining Your Category
At this point, you should have 6 subcategories. It is time to take a step back and conceptually summarize how these 6 subcategories make up your category. A causal condition leads to a phenomenon to occur. In reaction to this phenomenon, your research participant came up with strategies that led to consequences. All of which took place within some context and were influenced by intervening conditions.
All of this should be summarized in the description of your category, as you write this up in narrative form. Re-read your subcategories and supporting codes and quotes, and flesh out what you are seeing in one or more memos. As a final step give your category a short but ideally nuanced name. You will be utilizing all your categories in the next step of selective coding!
Example: Having done all this analysis and created all these subcategories, we can now confidently name our category emerging from our data:
“Socializing to Cope with Anxiety During Covid-19 Lockdown”
We then further summarize our findings as part of our category description:
In our grounded research, we found that Covid-19 induced Lockdown Loneliness (causal condition) led to our research participants experiencing Anxiety (phenomenon). There was some variety to the intensity and frequency that the research participants experienced that anxiety (context) with participants experiencing more intense bouts of anxiety being those who had experienced panic attacks pre Covid-19 (intervening circumstances). To address their anxiety they scheduled both in-person and virtual social hangouts (strategies). Whether people chose in-person and virtual hangouts was primarily due to what their current living situation was (context) and whether they had identified as introvert or extrovert (intervening circumstances). While these hangouts had some positive mental effects like breaking up the day and elation from laughing, they also had the negative side effects of zoom burnout and worries that in-person hanging out might lead to contracting Covid-19 (consequences).
Remember Grounded Theory and Axial Coding are Iterative!
Above we will be describing a step by step process of axial coding in research by applying this Coding Paradigm to find subcategories that we connect to develop a category. It should be noted that axial coding is an iterative process. Even though we broke it into steps, you should frequently go back and revisit earlier steps.
As you gain experience with axial coding in research, you might find that you no longer need to follow these steps so rigidly, and are able to analyze with all elements of the Coding Paradigm in mind. We have found these steps are a great way for novice researchers to become acquainted with axial coding.
Axial coding is a part of the process of grounded theory. Learn more about grounded theory in our Practical Guide to Grounded Theory.
Qualitative analysis doesn't have to be overwhelming
Take Delve's free online course to learn how to find themes and patterns in your qualitative data. Get started here.
Try Delve, Axial Coding Software
Online software such as Delve can help streamline how you’re coding your qualitative coding. Try a free trial or request a demo of the Delve.
Reference
Corbin, J., & Strauss, A. (1990). Grounded theory research: Procedures, canons, and evaluative criteria. Qualitative Sociology, 13, 3-21.
Saldana, Johnny (2009). The Coding Manual for Qualitative Researchers. Thousand Oaks, California: Sage.
