Introduction to Qualitative Coding and Content Analysis

Philosophical Foundations of Qualitative Research

• Big Theory refers to the broader philosophical positions and assumptions that underpin all scientific research.

• Ontology concerns assumptions about the nature of reality (what exists and what can be known about it).

• Epistemology concerns assumptions about knowledge, specifically how knowledge is acquired and validated.

• Ontological and epistemological assumptions direct research questions, data collection methods, analytical strategies, and overall data interpretation.

• Qualitative methods and qualitative data can be utilized across both qualitative and quantitative paradigms.

• The qualitative spectrum (small q vs. Big Q):

  • small q: Uses qualitative data collection tools or data sets within a positivist or post-positivist framework (aiming for objective measurement, standardized coding, and minimizing researcher bias).

  • Big Q: Fully embraces a qualitative philosophical worldview, emphasizing constructivism, relativism, researcher subjectivity, and the co-construction of meaning.

  • Many qualitative research projects occupy positions along the continuum between small q and Big Q.

Fundamentals of Qualitative Coding

• Coding is the analytical process of making sense of qualitative data by assigning descriptive or conceptual labels to meaningful segments of text or media.

• Code: A label capturing an analytically relevant idea, concept, experience, or meaning contained within a specific data segment.

• Purpose of Coding:

  • Organizes unstructured qualitative data relative to the primary research questions.

  • Renders large qualitative datasets manageable for systematic analysis.

  • Enables comparisons across the dataset to uncover patterns, connections, and structures that can be developed into broader categories or themes.

• Coding serves as a foundational analytical technique across multiple qualitative methods, including Content Analysis, Thematic Analysis, and Grounded Theory.

Types of Codes: Explicit vs. Implicit Continuum

• Codes exist along a continuum ranging from manifest/semantic content to latent/implicit meanings.

• Manifest / Semantic (Explicit) Coding:

  • Focuses directly on what the participant explicitly states or what is directly observable in the text.

  • Minimizes researcher interpretation to preserve literal meaning.

  • Predominantly aligned with realist and post-positivist paradigms (small q) where the objective is identifying a stable truth independent of the researcher.

• Latent / Implicit Coding:

  • Captures underlying meanings, implicit assumptions, sociological contexts, and psychological dynamics.

  • Requires active interpretation by the researcher.

  • Predominantly aligned with relativist, constructivist, and participatory paradigms (Big Q) where meanings are understood as context-dependent and co-constructed.

• Example Analysis:

  • Research Question: What are young people's experiences of seeking professional help for mental health difficulties?

  • Participant Statement (Dave): "I kept putting it off because I thought my problems weren't serious enough. I felt like I would be wasting someone's time."

  • Explicit/Semantic Codes:

  • Putting off help-seeking

  • Perceiving problems as not serious enough

  • Concern about wasting professionals' time

  • Implicit/Latent Codes:

  • Feeling undeserving of care

  • Being dismissive of own struggles

  • Feeling a burden to others

Comparative Exemplars of Explicit and Implicit Coding

• Excerpt 1: "Once I finally spoke to my GP, I actually felt relieved. Just saying it out loud to someone made it feel more manageable."

  • Explicit Codes:

  • Speaking to a GP

  • Feeling relief after disclosure

  • Talking makes difficulties feel more manageable

  • Implicit Codes:

  • Feeling lighter after disclosing

  • Disclosing as a turning point

  • Relief from sharing the burden with someone else

• Excerpt 2: "My friends were the ones who convinced me to get help. I don't think I would have done it without them."

  • Explicit Codes:

  • Friends encouraging help-seeking

  • Needing encouragement from others

  • Implicit Codes:

  • Friends validating the need to seek help

  • Social permission to seek help

• Excerpt 3: "The hardest part was having to explain everything again to different people. It made me feel like nobody really knew what was going on."

  • Explicit Codes:

  • Having to disclose multiple times

  • Feeling that professionals did not know what was happening

  • Implicit Codes:

  • Lack of understanding from professional staff

  • Dysfunctional healthcare system

• Excerpt 4: "I was nervous before the appointment because I didn't know what they were going to ask me or whether they would take me seriously."

  • Explicit Codes:

  • Feeling nervous before an appointment

  • Uncertainty about what would happen

  • Concern about not being taken seriously

  • Implicit Codes:

  • Negative feelings due to lack of trust in professional services

  • Services not communicating what to expect

Coding Orientations: Deductive, Inductive, and Mixed Approaches

• Qualitative coding can be approached deductively, inductively, or via a combined mixed orientation.

• Deductive Orientation (Theory-Driven):

• Codes are derived from pre-existing theories, theoretical concepts, or standardized frameworks before or during data analysis.

• Used to test, extend, or apply existing conceptual models to a new dataset.

• Example (Research Question: How do LGBTQ+ university students experience expressing their identities on campus?):

  • Alex: "People often assume I'm straight unless I tell them otherwise. It's not always offensive, but it gets tiring having to correct people all the time."

    • Deductive Codes: Heteronormativity (Warner, 1991); Gender as a binary (Butler, 1990).

  • Sunny: "I'm always thinking about how I dress depending on where I'm going. Sometimes the way I look, or how other people see me, doesn't match how I feel inside, and that can make me feel really uncomfortable."

    • Deductive Codes: Gender performativity (Butler, 1990); Gender dysmorphia (Cooper et al., 2020).

  • Mo: "Sometimes people treat me differently and I'm never sure why. I don't know if it's because I'm queer, because I'm Black, or maybe both."

    • Deductive Codes: Intersectionality (Crenshaw, 1989).

  • Inductive Orientation (Data-Driven):

• Codes are generated directly from patterns, meanings, and structural features present within the raw data.

• No predetermined coding frame is forced onto the data.

• Example (Research Question: How do LGBTQ+ university students experience expressing their identities on campus?):

  • Alex: "People often assume I'm straight…"

    • Inductive Codes: Being assumed to be straight; Frustration with others' gender assumption; Emotional burden of explaining identity.

  • Sunny: "I'm always thinking about how I dress…"

    • Inductive Codes: Changing gender expression across contexts; Doubts about identity; Mismatch between outward appearance and inner sense of self.

  • Mo: "Sometimes people treat me differently…"

    • Inductive Codes: Feeling like they are treated differently; Uncertainty about the reason for discrimination; Difficulty separating racism from queerphobia.

• From a Big Q perspective, pure inductive coding is impossible because a researcher's existing knowledge and subjective lens inevitably shape their analytical gaze.

  • Mixed Approach Orientation:

• Combines theory-driven and data-driven coding within a single analysis.

• Researchers may start with a core set of deductive codes based on literature, while leaving space to add new inductive codes as unexpected topics emerge.

• Alternatively, initial coding may be conducted inductively, followed by mapping emerging categories to existing theoretical concepts.

• Widely utilized in qualitative psychological research to balance empirical openness with theoretical groundings.

Transitioning from Codes to Categories

• Structure of Categorization:

  • Codes represent the smallest analytical unit (labels assigned to text segments).

  • Leaving codes isolated results in fragmented analysis.

  • Analysts combine, compare, and organize related codes into broader categories (also termed meta-codes or dimensions).

  • Categories may subsequently be abstracted into overarching themes depending on the chosen methodology.

• Levels of Interpretative Abstraction:

  • Low Interpretative Level (Manifest/Concrete):

  • Codes: Apple, Pear, Orange $\rightarrow$ Possible Category: Fruit / Food.

  • Codes: Bus, Train, Bicycle $\rightarrow$ Possible Category: Modes of Transport / Vehicles.

  • Moderate to High Interpretative Level (Latent/Conceptual):

  • Codes: Sadness, Anger, Shame $\rightarrow$ Possible Category: Negative Affective States / Emotional Distress.

• Applied Category Grouping (Research Question: How do university students experience academic stress?):

  • Example 1 (Somatic Indicators):

  • Code: Difficulty sleeping before deadlines ("I struggle to fall asleep before deadlines.")

  • Code: Stress-related headaches ("I get headaches when I have lots of assignments")

  • Code: Racing heart during stress ("Sometimes my heart feels like it's racing.")

  • Possible Meta-Category: Physical symptoms of stress / Somatic responses.

  • Example 2 (Avoidance Dynamics):

  • Code: Delaying work due to fear of failure ("I keep putting assignments off because I'm scared I won't do them well.")

  • Code: Avoiding academic tasks due to anxiety ("Sometimes I avoid even opening Canvas because it makes me anxious.")

  • Code: Feeling unable to begin when overwhelmed ("The more overwhelmed I feel, the less I seem able to start.")

  • Possible Meta-Category: Academic avoidance behaviors / Procrastination driven by anxiety.

  • Example 3 (Self-Appraisal Dynamics):

  • Code: Comparing oneself negatively with peers ("Everyone else seems to know what they're doing but me.")

  • Code: Questioning academic belonging ("When I get a low mark, I start wondering whether I belong at university.")

  • Code: Fear that others overestimate one's ability ("I feel like eventually people are going to realise I'm not as good as they think.")

  • Possible Meta-Category: Imposter phenomenon / Academic belonging insecurity.

Enhancing Validity, Reliability, and Rigor in Coding

• As qualitative data grow more complex, interpretative variance increases.

• Validity:

• Refers to the accuracy and meaningfulness of coding.

• Evaluates whether codes accurately represent the core concept, experience, or feature of the data they are intended to capture.

  • Reliability:

• Refers to consistency across coding applications.

• Evaluates whether the same data would be coded similarly by different researchers or by the same researcher across different time points.

  • Coding Frame (Codebook):

• An analytical document used to structure and systematically apply codes across qualitative datasets.

• Contains explicit code names, operational definitions, inclusion and exclusion criteria, and illustrative examples.

• Organizes sub-codes underneath umbrella meta-codes (categories).

• Can be refined iteratively throughout data analysis as overlap or gaps emerge.

• Exemplar Coding Frame Structure (Study on coping mechanisms post-tsunami):

  • Meta-code: Using prayer as a coping mechanism

    • Code: Praying for emotional regulation | Definition: References to praying as a way to stay calm, ease anxiety, and fear | Example: "We pray to Allah to eliminate the fear"

    • Code: Praying to prevent further disasters | Definition: References to praying to God for no more disasters | Example: "We have to keep praying and have faith that God won't send us any more earthquakes"

    • Code: Praying as a way of social connection | Definition: References to praying in group. Excludes non-religious social connection | Example: "All the neighbours who were around there, we all gathered at the mosque, and we prayed together"

  • Inter-Coder Reliability:

• Assesses the level of coding agreement between two or more independent researchers coding an identical data subset using the same coding frame.

• Metrics used include percentage agreement and Cohen's kappa coefficient ($\kappa$).

• Discrepancies are used constructively to resolve ambiguities, tighten code definitions, and remove overlap.

• Inter-coder reliability is utilized strictly in small q or post-positivist qualitative research; radical Big Q qualitative research rejects inter-coder reliability checks as philosophical mismatches with subjectivity.

Content Analysis Framework

• Content Analysis: A systematic analytical methodology used to categorize features of textual or visual content and measure their presence or occurrence frequencies.

• Frequently situated within small q or transitional small q / Big Q paradigms.

• Operates by quantifying category occurrences to determine salience (relative importance or prominence of specific ideas relative to research questions).

• Focuses on meaning units (phrases, sentences, or explicit ideas) rather than bare word-frequency counts.

• The 10-Step Content Analysis Process (Joffe & Yardley, 2004):

1. Define the research question: Determine what specific insights the study seeks to generate.

2. Familiarise yourself with the material: Read and view data closely to map broad content.

3. Decide the unit of coding: Define the boundaries of analysis (e.g., word, phrase, sentence, meaning unit, image, or full text).

4. Develop a coding frame: Build codes deductively, inductively, or via a mixed orientation.

5. Define and refine codes: Assign precise definitions, operational boundaries, and concrete examples to all codes; organize codes under meta-codes.

6. Check coding reliability (if appropriate): Deploy independent coders and revise code definitions based on inter-coder agreement metrics.

7. Code the full dataset: Apply the finalized coding frame systematically across all data.

8. Analyse patterns in the codes: Examine frequencies, cross-group differences, co-occurrences, and structural relationships.

9. Interpret the findings: Analyze code distribution patterns within broad theoretical contexts and existing academic literature.

10. Report the analysis: Document category generation procedures, present quantitative distributions/patterns, and supply illustrative qualitative quotes.

Empirical Case Studies in Content Analysis

• Case Study 1: Small q Post-Positivist Paradigm (Moreno et al., 2011):

  • Title: Feeling bad on Facebook: depression disclosures by college students on a social networking site.

  • Objective: Measure the prevalence and characteristics of depression symptom disclosures on Facebook profiles among college students and evaluate alignment with Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria for Major Depressive Episode (MDE).

  • Hypothesis: Stigma concerning public disclosures of mental health issues would result in very low symptom prevalence.

  • Sample: 200 public Facebook profiles belonging to undergraduate students ($M_{age} = 20\,\text{years}$).

  • Data: Profile status updates posted over a continuous 12-month period.

  • Method: Systematic content analysis coding updates against DSM-IV criteria.

  • Findings:

  • $25\%$ of profiles contained posts meeting criteria for depressive symptoms.

  • $2.5\%$ of profiles met coding criteria for a Major Depressive Episode (MDE).

  • Social reinforcement (e.g., supportive friend comments) correlated with increased disclosure frequency.

  • DSM-IV Criteria Content Analysis Coding Map:

- *Depressed mood* (Key phrases: Sad, empty, crying, tearful) $\rightarrow$ Excerpt: "Tom is pretty sad. Time for some whiny music...thank god for that"; "Mary has tears in her eyes"

- *Decreased interest/pleasure* (Key phrases: Not having fun, don't feel like doing anything) $\rightarrow$ Excerpt: "Jane doesn't feel like getting up today, or doing anything"

- *Appetite changes* (Key phrases: No appetite, can't stop eating) $\rightarrow$ Excerpt: "Amy has no appetite right now"

- *Sleep problems* (Key phrases: Sleeping too much, slept $> 10\,\text{hr}$, fatigue) $\rightarrow$ Excerpt: "Ann needs to stop being lame and so tired so that she can go out and socialize more..."

- *Psychomotor agitation/retardation* (Key phrases: Feeling slow) $\rightarrow$ Excerpt: "Jeff is moving bogged down"

- *Loss of energy* (Key phrases: Can't get motivated, no drive) $\rightarrow$ Excerpt: "Joe has lost his motivation"; "Mia has no drive"

- *Guilt/Worthlessness* (Key phrases: Feel guilty, "I am stupid") $\rightarrow$ Excerpt: "Matt feels absolutely useless..."; "Kate hates herself right now"

- *Indecisiveness* (Key phrases: Can't decide, can't make up mind) $\rightarrow$ Excerpt: "Jim is frustrated and indecisive....argh."

- *Suicidal ideation* (Key phrases: Thoughts of death, jumping) $\rightarrow$ Excerpt: None observed in dataset.

• Case Study 2: Constructivist / Interpretative Content Analysis (Joffe & Haarhoff, 2002):

  • Title: Representations of far-flung illnesses: The case of Ebola in Britain.

  • Objective: Explore media representations and public understanding of Ebola in Britain.

  • Data: British newspaper articles (tabloids and broadsheets) coupled with semi-structured reader interviews.

  • Method: Content analysis using an inductively driven coding frame, informed conceptually by Social Representations Theory.

  • Findings: The analysis extended beyond frequency counts to uncover symbolic meaning. Ebola was systematically linked in British media and public discourse with Africa, poverty, tribal ritual, and "otherness," providing British citizens with a psychological mechanism to distance themselves from the threat of disease.

Execution Guidelines for Content Analysis Projects

• Analytical Boundaries and Artificial Intelligence:

  • Automated AI tools may align with raw manifest content quantification in pure small q approaches.

  • When analytical tasks require qualitative interpretative judgment, researcher subjectivity, and context evaluation, automated AI coding is methodologically inappropriate.

• Practical Content Analysis Workflow (Healthcare Patient Satisfaction Example):

  • Dataset: Publicly available feedback comments and rating evaluations for National Health Service (NHS) GP practices in England.

  • Core Research Questions:

  1. What factors increase patients' satisfaction with their GP surgery?

  2. What factors decrease patients' satisfaction with their GP surgery?

  • Procedural Execution:

  1. Familiarization: Annotate raw text comments to identify recurring operational patterns (e.g., appointment availability, staff empathy, wait times).

  2. Coding Frame Construction: Construct hierarchical meta-codes and sub-codes equipped with definitions and clear criteria.

  3. Systematic Coding: Apply code units across all feedback entries and calculate frequencies to identify dominant drivers of patient satisfaction.