---
name: qualitative-analysis
description: Analyze interviews, observations, documents, and other qualitative materials with a coherent methodological stance, contextual interpretation, reflexivity, and traceable evidence.
---

# Qualitative analysis

Use this Academic Research Assistant skill for qualitative interpretation and analytic development. Follow the [research contract](../../docs/RESEARCH-CONTRACT.md). Select an approach that fits the question and materials; thematic analysis, grounded theory, discourse analysis, narrative analysis, and case analysis are not interchangeable coding templates.

## Inputs and analytic stance

Obtain the question, context, sampling and data-generation history, source materials, consent and reuse limits, disciplinary approach, and researcher position. Determine whether the task is an initial exploration, coding support, a focused reanalysis, or development of an interpretive account. If only selected excerpts are supplied, limit conclusions to those excerpts and do not claim coverage of the full dataset.

Clarify the intended relationship to theory and interpretation. A descriptive summary, an account of participants' meanings, an analysis of language in use, and theory development ask different things of the same text. Establish whether codes are primarily inductive, theory-informed, deductive, or deliberately combined, and describe how this choice affects what becomes visible.

## Analysis operations

1. Inventory the corpus and provenance. Preserve participant or case identifiers, source type, date, context, and transcript version without exposing identifying details. Review transcription conventions and any translation because pauses, wording, and context may matter analytically.
2. Read across the material for familiarization. Write short memos about emerging questions, context, surprising passages, and the analyst's assumptions. Do not let an automatically generated summary replace engagement with the underlying material.
3. Develop codes or analytic annotations appropriate to the method. Distinguish a participant's wording from an analyst's concept. If using a codebook, define codes, boundaries, and examples from authorized materials, and record revisions. Do not force passages into mutually exclusive codes when overlap is meaningful.
4. Compare within and across cases. Look for variation, exceptions, contradictions, silences, temporal development, and links to context. Repeated wording is not necessarily the most important finding; a rare account can expose a crucial boundary or mechanism.
5. Build candidate themes, categories, narratives, or interpretive propositions. Explain the organizing idea and its relationship to the research question. A topic heading such as “barriers” is not yet an analytic theme without an account of what the barriers mean or how they operate.
6. Return to the source materials to test the interpretation. Seek disconfirming cases, distinguish competing readings, and revise claims that depend on decontextualized excerpts. Keep a trail from the proposed finding through analytic memos to supporting material.
7. Draft an account that connects interpretation, evidence, context, and limits. Select excerpts that substantiate an analytic point; do not decorate a predetermined argument with convenient quotations.

## Quality and method boundaries

Use quality practices consistent with the chosen approach. Reflexive work may emphasize interpretive transparency and researcher positioning; a structured team-coding study may need coder training and agreement procedures. Inter-rater agreement, participant checking, and saturation are not universal quality requirements. If used, explain what they assess and what they cannot establish.

Do not convert the frequency of coded passages into participant prevalence without a justified counting unit and sampling rationale. Do not describe a small purposive sample as statistically representative. Transferability depends on contextual description and a reasoned comparison, not an automatic population claim.

Treat machine-generated codes or summaries as suggestions for researcher review. Disclose consequential assistance and preserve human responsibility for contextual interpretation. Sensitive or culturally specific material may require expertise that the assistant lacks. For mixed-method projects, define how qualitative and quantitative findings will inform one another rather than using quotations merely to validate a statistical result.

## Deliverable and verification

Deliver the analytic approach, corpus description, coding or annotation structure where relevant, memos, findings with traceable excerpts, contrary cases, and reflexive limitations. Verify quoted text against the source and anonymize according to actual agreements without changing its meaning. Label translations and analyst reconstructions. Route questions of design to [methodology design](../methodology-design/SKILL.md) and synthesis of multiple qualitative studies to [systematic review](../systematic-review/SKILL.md). Do not claim that an interpretation is exhaustive merely because all files were processed.