What Single-Axis Coding Misses in STEM Interviews
Qualitative STEM research has often treated race, gender, and class as separate variables. That structure makes a transcript easier to sort. It also breaks apart accounts in which those forces operate at the same moment. When a participant describes being passed over for a game-design task, what is lost if each identity receives a separate code and the interaction receives none?
Consider a student whose design proposal is dismissed as too ambitious while classmates receive encouragement for similar work. A transcript may contain references to money, gendered expectations, and assumptions about technical ability. Filing each reference under a demographic heading preserves the words but obscures the event: who could claim expertise, whose proposal sounded credible, and what resources made participation possible.
The foundational 1989 framing of intersectionality offers a sharper unit of analysis. It asks researchers to examine what happens where systems of power meet. For interview coding, that means identifying the institutional action within an account before deciding which identities matter to its interpretation.
Build Codes Around Overlapping Identities
Start with the research question and a small set of transcripts. Mark passages where participants explicitly connect an identity to an opportunity, constraint, or judgment in a STEM setting. Then draft codes around those interactions. “Race,” “gender,” and “class” can remain useful descriptors; experiential codes explain what happened.
Define the action each code captures
A workable entry names the pattern, defines its boundaries, and tells coders what evidence to seek. The table below illustrates a codebook structure, rather than a universal set of categories.
| Code name | Definition | Include when | Exclude when |
|---|---|---|---|
| Contribution Overlooked | A participant links the dismissal of technical work to interacting identity-based expectations. | The account connects an identity-related assumption with a specific ignored contribution. | A contribution is ignored, but the participant supplies no basis for an identity-linked interpretation. |
| Resource-Gated Participation | Access to time, equipment, or support shapes how a participant can meet expectations tied to identity. | The participant describes both a resource constraint and the social judgment attached to it. | The account describes a resource constraint without that connection. |
Inclusion and exclusion criteria keep a coder from treating every mention of identity as evidence of an intersectional mechanism. An excerpt can receive more than one code when the participant describes distinct processes. The memo should explain how those processes meet in the particular institution: a classroom, lab, competition, or design team.
Established intersectional frameworks can guide the questions asked of a transcript, but predefined codes have a narrow reach. This approach is best suited to qualitative accounts in which participants articulate overlapping barriers. A highly structured survey response may leave too little context to support the same interpretation. Leave room in the codebook for an emerging code when a participant names a mechanism the initial definitions miss.
Move From Participant Language to Patterns
A game-design interview provides a useful walkthrough. The following excerpt is constructed for illustration; it is not a participant quotation:
“When I proposed the character system, they asked if I could handle the coding. I was also working evenings, so I missed the optional lab where everyone made those decisions.”
Keep the first pass close to the transcript
On the first cycle, mark “asked if I could handle the coding” and “missed the optional lab” as separate In Vivo phrases. Add a descriptive code such as technical ability questioned to the first span and decision-making access constrained to the second. Record the speaker’s explanation of evening work without inferring a reason the participant has not given.
Now read the surrounding exchange. Did the participant connect the question about coding ability to race, gender, or both? Did they describe evening work as an economic constraint? If the transcript does not make those connections, keep the codes provisional. Intersectional analysis needs the participant’s account and its context, not a demographic profile used to fill gaps.
Group only what the evidence supports
In a second cycle, compare first-pass codes across excerpts. Where participants describe identity-linked doubts about competence alongside restricted access to informal decision-making, a focused code such as conditional technical belonging may capture the shared pattern. Retain the initial phrases beneath it. They show whether the broader theme reflects what participants said or merely what the team expected to find.
This sequence also makes a useful audit trail: transcript span, first-cycle phrase, focused code, and written rationale. A reader of the research & publications record should be able to follow that path back to the account.
Use Memos to Keep Institutions in View
A memo records the interpretive step between an excerpt and a claim. Without one, “conditional technical belonging” can become an attractive label detached from the lab rules, team dynamics, and participant language that gave it meaning.
Write a memo after a meaningful coding session, while the uncertain decisions are still fresh. Use a repeatable sequence:
- Anchor the claim. Identify the transcript passage and the participant’s exact words.
- Name the setting. Specify the assignment, gatekeeping decision, informal meeting, or resource constraint described.
- Trace the interaction. Explain which identity-linked expectations appear together and how the participant connects them.
- Test another reading. Note what evidence would favor a different code or leave the interpretation unresolved.
- Locate the researcher. Record assumptions the coder brings to the passage and questions for the next interview or team discussion.
The positionality entry deserves plain language. A coder who has enjoyed easy access to optional labs might initially read attendance as motivation. Another might notice that the meeting schedule selects for students with free evenings. Neither reading should replace the participant’s explanation; both can expose a question worth checking against the transcript. Memos make that decision visible instead of burying it in a final theme name.
Resolve Coding Disagreements Without Erasing Them
In a simple arrangement, two coders can read the same passage accurately and still disagree about its mechanism. One marks a challenge to technical competence; the other sees restricted access to the room where competence is recognized. Their different positionalities may help reveal how those processes reinforce one another.
Begin a consensus discussion with the excerpt, its surrounding exchange, and each coder’s written rationale. Ask each person to identify the participant language that supports their code. Then compare the codebook’s inclusion and exclusion criteria. If both readings have support, preserve both codes and add a memo explaining their relationship. If neither meets the criteria, revise the definition or leave the passage unresolved.
Reliability measures can flag where coders diverge, but agreement alone cannot establish that an intersectional reading is sound. A majority decision can also silence a well-supported minority interpretation. Keep a discrepancy log that records the competing readings, the evidence considered, and the reason for the final decision. Revisit disputed passages when a later interview clarifies the mechanism.
Crenshaw’s 1989 traffic-intersection metaphor locates the crucial evidence at the crossing, where more than one path can contribute to the same injury.
