Skip to content
Back to search results

Axial Coding — Relational Category Development in Qualitative Research

Introduced by
Anselm Strauss and Juliet Corbin
Year
1990
Type
Qualitative research method
Family
Qualitative Coding
Font size

What does it do?

Axial coding is the second major analytical step in grounded theory analysis, performed after open coding.

Introduced by Anselm Strauss and Juliet Corbin in 1990, it involves systematically re-examining and reorganising the many discrete codes generated during open coding by identifying a central (axial) category and mapping the causal conditions, contextual factors, intervening conditions, action-interaction strategies, and consequences that surround it. The goal is to move from a fragmented list of codes to a coherent relational structure that reflects how concepts interconnect in the data.

The intuition behind the method

Think of open coding as disassembling a watch into its smallest parts — springs, gears, hands, screws. Axial coding is the step where you figure out how those parts fit together: which parts drive which other parts, what conditions must be in place for the mechanism to work, and what happens when it does or does not. In qualitative data terms, you identify one key phenomenon (the axis) and then ask: what causes it, who is involved, under what conditions, how do people respond, and with what outcomes? This transforms a collection of isolated labels into a connected explanatory picture.

History

Axial coding was formally introduced by Anselm Strauss and Juliet Corbin in their landmark 1990 textbook Basics of Qualitative Research, which systematised the grounded theory methodology that Strauss had originally developed with Barney Glaser in the 1960s. Glaser, who favoured a more emergent, less structured approach, later criticised the Strauss-Corbin coding paradigm as overly prescriptive. Kathy Charmaz's constructivist grounded theory (2006) offered a further critique, advocating for flexible, iterative coding without a fixed paradigm structure. Today axial coding remains central to the Straussian tradition of grounded theory and is widely taught in health sciences, education, and social science research methods courses.

When to use it

Axial coding is appropriate as part of a grounded theory study when the researcher has already generated a substantial body of open codes and is ready to move beyond description toward explanation. It is best suited to research questions that ask how or why a social process or phenomenon unfolds — not simply what it looks like. The method requires rich, detailed qualitative data (typically 20 or more in-depth interviews, supplemented by field notes or documents) and is inappropriate as a standalone technique: it is analytically dependent on a preceding open coding phase. Axial coding is not recommended when the research goal is purely descriptive, when the dataset is small, or when a lighter-weight approach such as thematic analysis would suffice.

Limitations and considerations

  • Analytically demanding — requires a researcher already skilled in open coding, memo-writing, and theoretical sensitivity before axial coding can be applied productively.
  • The Strauss-Corbin coding paradigm has been criticised (notably by Charmaz) as imposing a predetermined conceptual structure on data rather than allowing categories to emerge purely from the material.
  • Not appropriate for small datasets; without sufficient data diversity the relational map will be under-populated and speculative.
  • The output is a relational model, not a finished theory — further selective coding is required to reach a complete grounded theory.
  • Time-intensive: producing a defensible axial coding map typically requires multiple cycles of coding, memo-writing, and data return.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Moves qualitative analysis beyond simple description by surfacing causal, contextual, and consequential relationships among categories.
  • Provides a systematic and transparent analytical framework (the coding paradigm) that makes the reasoning behind category relationships auditable.
  • Produces a relational structure that can be refined into a formal grounded theory in the subsequent selective coding phase.
  • Well-suited to understanding complex social processes where multiple conditions interact to shape outcomes.
  • Encourages constant comparison between emerging categories and raw data, keeping the analysis empirically grounded throughout.

Applications

  • Developing a grounded theory of how nurses manage moral distress in intensive care settings — mapping causal conditions, coping strategies, and outcomes.
  • Analysing how entrepreneurs make sense of business failure — identifying contextual factors, intervening conditions, and the role of social support.
  • Building a relational model of student disengagement in online learning environments, linking instructional conditions to behavioural responses.
  • Investigating how community members adapt to environmental displacement — tracing the conditions, strategies, and long-term consequences across interview data.
  • Studying how healthcare teams coordinate care across professional boundaries — identifying the conditions under which collaborative strategies succeed or break down.
  • Examining patient experiences of diagnostic uncertainty, mapping the causal chain from symptom onset through interpretive strategies to health-seeking behaviour.

How it works

How it works

1

Begin by reviewing all open codes and initial memos generated in the open coding phase. Look for codes that appear frequently, carry strong conceptual weight, or seem to be pivot points around which other codes cluster. Group closely related open codes into broader conceptual categories, giving each category a descriptive label that captures its analytic meaning rather than simply restating a word from the data.

2

Select the central category — the concept that appears most frequently, is most linked to other categories, and best accounts for the variation in the data. This is the 'axis' of axial coding. Additional subcategories are then organised around this central category, each representing a distinct dimension or property of the phenomenon being studied.

3

For each category-subcategory pairing, articulate the nature of the relationship: is the subcategory a causal condition that gives rise to the central phenomenon, a contextual factor that shapes how it unfolds, an intervening condition that moderates the response, or a consequence of the phenomenon? Writing these relationships out in propositional form ('When X occurs in context Y, actors respond with strategy Z, leading to outcome W') makes the emerging theory testable and transparent.

4

Apply the coding paradigm proposed by Strauss and Corbin — a structured analytical framework that organises the relational map into six slots: causal conditions, the phenomenon itself, context, intervening conditions, action-interaction strategies, and consequences. Not every dataset will populate all six slots equally, but working through each systematically ensures no major relational dimension is overlooked.

5

Return to the raw data — transcripts, field notes, or documents — to verify that the proposed category relationships are grounded in specific evidence. Negative cases (instances that do not fit the emerging relational map) are especially important: they may reveal conditional boundaries of the categories or prompt refinement of the central category itself. Memo-writing at this stage captures the reasoning behind each relationship.

Common pitfalls

  • Applying axial coding before open coding is complete — premature closure around a central category too early in analysis forces data into predetermined slots and suppresses disconfirming evidence.
  • Treating the coding paradigm as a rigid checklist — not every dataset will exhibit all six paradigm elements equally; forcing the data into slots that are not warranted by the evidence weakens validity.
  • Ignoring negative or deviant cases — instances that do not fit the emerging relational map are analytically valuable; setting them aside rather than using them to refine category boundaries is a common and costly error.
  • Conflating axial coding with thematic analysis — axial coding is specifically designed to build relational models of social processes, not merely to identify recurring themes; using the label without the relational logic misrepresents the method.
  • Losing connection to the raw data — as the analysis becomes more abstract, researchers sometimes stop returning to transcripts to validate relationship claims; every proposed relationship should be traceable to specific data excerpts.

Frequently asked questions

Is axial coding the same as thematic analysis?

No. Thematic analysis identifies recurring patterns (themes) across a dataset and is primarily descriptive. Axial coding, by contrast, is an explicitly relational procedure: it seeks to explain how and why a central phenomenon occurs by mapping causal conditions, contextual factors, intervening conditions, strategies, and consequences. Axial coding belongs to the grounded theory tradition and is analytically more demanding than thematic analysis.

Can I use axial coding without the rest of grounded theory?

Axial coding is designed as the second phase of grounded theory analysis, following open coding. It relies on the open codes and memos produced in that first phase. Using the term 'axial coding' for a standalone relational analysis is technically a misuse of the label. If you want to map relationships among qualitative categories outside a grounded theory framework, approaches such as framework analysis or systems mapping may be more appropriate.

What is the difference between the Strauss-Corbin and Charmaz approaches to coding?

Strauss and Corbin advocate a structured coding paradigm with defined slots (causal conditions, phenomenon, context, intervening conditions, strategies, consequences). Charmaz's constructivist grounded theory rejects this paradigm as too prescriptive, preferring open-ended, iterative coding that allows conceptual categories to emerge more freely. Strauss-Corbin suits researchers who value systematic structure; Charmaz suits those who prioritise interpretive flexibility and are concerned about imposing categories on data.

How do I know when axial coding is finished?

Axial coding is sufficiently complete when every major category in the data has been related to a central category, the relationships among categories are supported by specific data excerpts, negative cases have been examined and either accommodated or explained, and the relational map is stable across repeated returns to the data. The next step is selective coding, which integrates the categories around a core category into a formal grounded theory statement.

What software can support axial coding?

ATLAS.ti, NVivo, and MAXQDA all support axial coding workflows: they allow researchers to create hierarchical code families, link codes with relationship connectors, and visualise category maps. None of the software performs the analytical work — assigning relational roles to categories and validating them against data excerpts is the researcher's intellectual task. Software reduces the administrative burden of managing large code sets and memos.

Related methods

Sources

  1. Strauss, A., & Corbin, J. (1990). Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Sage.

    ISBN: 978-0803932456
    Copied.
  2. Charmaz, K. (2006). Constructing Grounded Theory: A Practical Guide Through Qualitative Analysis. Sage.

    Open publication reference
    Copied.