> Qualitative Analysis
Choosing Your Path: Codebook or RTA?
When to use codebook / coding-reliability vs reflexive thematic analysis — with concrete domain examples and a decision walkthrough. · 12 min
The Overview introduced two traditions — codebook / coding-reliability and reflexive thematic analysis (RTA). Before you start coding, decide which one your study calls for: the choice shapes how you code, whether you measure agreement, and how you write up rigor. This lesson compares them, shows who does what, and helps you make that call up front. (Later, Comparing the Two Paths shows the same data analyzed both ways, side by side, as the payoff.)
What pulls you toward each
| Dimension | Codebook / coding-reliability | Reflexive TA |
|---|---|---|
| Research question | Prevalence, comparison, “how often / which category” | Meaning, experience, “how / why” |
| Stance toward codes | Codes can be defined in advance; meaning is stable enough to share | Codes are interpretive; the analyst is the instrument |
| Who analyzes & why | Two+ coders who converge on a shared codebook | Two+ analysts who stay divergent — they discuss to deepen, never to score |
| Reliability claim | Wanted/expected (κ or α reported) | Rejected as a quality measure |
| Main payoff | Transferability + a reproducibility claim | Interpretive depth + theory-building |
| Reporting norms (within education, HCI & social science) | Venues/subfields that expect a reported reliability statistic, prevalence claims, or a reusable coding scheme | Venues/subfields that expect reflexivity, an interpretive stance, and rich situated accounts |
| Poor fit when… | The question is about contested meaning | You need prevalence counts or an auditor wants reproducibility |
Concrete examples
Both traditions are used heavily across education, HCI, and the social sciences — the choice is about your study, not your discipline. The same lab, in the same field, picks differently from project to project:
Who does what, and when
The table tells you which path; this shows what each looks like in practice. The confusing part isn’t the steps — it’s the people. Both paths can use two or more researchers, so the real question is what the second person is for. In the codebook path the two coders converge — they reconcile toward one shared instrument. In RTA they stay divergent — a critical friend challenges the reading to deepen it, and the codes are never merged or scored. Same two people, opposite purpose. Pick a path and step through it:
Both coders read every transcript and write their own memos and jottings — separately, so first impressions are not anchored to one another.
“I sat there for probably an hour before I asked”
Jotting: a whole hour stuck before asking — why wait? Social risk? Flag to watch.
Reflexive thematic analysis, same step: Both read + memo reflexively, separately.
Each coder drafts their own set of codes on the same few interviews, independently — two parallel first codebooks, not one. The overlap (in different words) is exactly what the next step resolves.
Coder A's draft
Coder B's draft
Reflexive thematic analysis, same step: Each codes for semantic + latent meaning.
The two compare their draft codebooks and reconcile them into one shared codebook. Disagreements are resolved by sharpening definitions — not by seniority or averaging.
Both coders' wording for the same affective barrier → one shared, defined code.
Reflexive thematic analysis, same step: Critical friend challenges the reading — never merged or scored.
They apply the shared codebook to the next set of interviews, add or split codes as needed, and consolidate again — repeating until the codebook stabilizes (constant comparison).
| Code | Definition |
|---|---|
| help-seeking delay | Waiting a substantial time while stuck before seeking help from a TA, peer, or instructor. |
| fear of negative evaluation | Reluctance to act (esp. to ask for help) driven by worry about being judged incompetent. |
| improvised debugging strategy | Self-taught, unsystematic tactics for locating a bug (trial-and-error, scattered print statements). |
Reflexive thematic analysis, same step: Construct themes through discussion.
Rigor rests on negotiated consensus plus a documented audit trail. A reported κ/α is a separate claim: once you consolidate, the coders are no longer independent, so a clean coefficient needs a held-out independent pass.
“v0.3 — merged 'felt stupid' into 'fear of negative evaluation'”
Audit trail: who changed what, when, and why. The log is the rigor — not a coder rank.
Reflexive thematic analysis, same step: Reflexivity + audit trail; IRR rejected on principle.
The analyst and a critical friend read deeply and keep reflexive memos — interrogating their own assumptions and position, not just the data.
“didn't want the TA to think I was dumb”
Reflexive note: am I reading my own student-anxiety into this? What is *their* stake in looking competent?
Codebook / coding-reliability, same step: Both coders read + memo, separately.
Each codes the data for semantic and latent meaning, bringing their own interpretive lens. Two readings are expected — and welcome.
Analyst's reading
Critical friend's reading
Codebook / coding-reliability, same step: Each drafts their own codes, independently.
They talk. The critical friend challenges the reading — "what makes you see it that way?" — to deepen and complicate it. They do not merge their codes or score agreement.
“"the tool talks past me" — the error withholds access”
Critical friend: what makes you read it as *withholding* rather than just confusing? Push the interpretation further.
Codebook / coding-reliability, same step: Reconcile the two drafts into one shared codebook.
The analyst constructs themes through ongoing discussion. Themes are interpretive outputs owned by the analyst, enriched — not validated — by the dialogue.
Asking for help as social risk
Debugging as lonely trial-and-error
Codebook / coding-reliability, same step: Apply to the next set, expand, consolidate — loop.
Rigor rests on reflexivity, a clear audit trail, and rich theme definitions. Inter-rater reliability is rejected on principle — the second person is here for depth, not agreement.
“Reflexivity statement + audit trail”
No κ here — a second reader deepened the analysis, they didn't 'agree' with it.
Codebook / coding-reliability, same step: Consensus + audit trail (κ only via a held-out pass).
We show both paths with two people, but either can be done by one — solo RTA is common, and a lone researcher can still keep a codebook; what needs a second person is the reliability or consensus claim, not the analysis itself (Braun & Clarke, 2021).
Walk it through
The table and examples show which path fits; now run your own study through it:
The next lessons are familiarization and first-cycle coding — the mechanic is shared, but the coding stance already starts to diverge there. The paths then split fully — codebook & team coding, then reflexive coding — and meet again in Comparing the Two Paths.