What Is Emotion Coding?
Emotion coding is a qualitative coding method that labels the emotions participants experience, recall, or express within data, interview transcripts, open-ended survey responses, diary entries, or observational notes. Codes might include frustrated, relieved, anxious, hopeful, angry, or conflicted. Unlike sentiment analysis, which classifies text along a positive-negative spectrum, emotion coding identifies specific feelings, captures their intensity, and tracks how emotions shift throughout an experience. The method was formalized by Johnny Saldana and draws on psychology's affect theories to bring emotional dimensions into qualitative analysis.
Why Emotion Coding Matters
Most qualitative coding methods focus on what people think, say, or do. Emotion coding adds how they feel, a dimension that often drives behavior more powerfully than rational assessment. A customer might logically understand that switching software saves money, but if the migration process triggers anxiety, they'll stay put. Emotion codes surface these invisible drivers. They're also essential for experience design, healthcare research, and any study where emotional response shapes outcomes.
How Emotion Coding Works
Identifying Emotions in Data
Emotions appear in data in three ways:
Explicitly stated. The participant directly names their feeling: "I was furious," "It was such a relief." These are the easiest to code, use the participant's emotion word (or a standardized synonym from your codebook).
Implied through language. The participant doesn't name the emotion but their word choice signals it: "I couldn't believe they'd do that to a loyal customer" implies betrayal or indignation. Coding implied emotions requires careful interpretation grounded in context.
Inferred from behavior or tone. In video recordings or observation notes, you might note laughter, sighing, hesitation, or raised voices. Facial expressions and body language carry emotional content that transcript text alone misses.
Building an Emotion Codebook
Start with a reference framework. Common options include:
- Ekman's six basic emotions: happiness, sadness, anger, fear, disgust, surprise.
- Plutchik's wheel of emotions: eight primary emotions with varying intensities (e.g., annoyance → anger → rage).
- Custom frameworks tailored to your research context. A customer experience study might use: trust, frustration, delight, anxiety, relief, indifference, confusion.
Define each emotion code with clear criteria and example data segments. This is especially important when multiple researchers are coding, as emotion interpretation is inherently subjective.
The Coding Process
Step 1: Read through the data segment. Ask: What emotion is the participant experiencing or expressing here?
Step 2: Assign the emotion code. If the emotion is explicitly stated, code accordingly. If implied or inferred, assign the code and document your reasoning in a memo.
Step 3: Note intensity. Many researchers add an intensity dimension, high, medium, low, or use Plutchik's graduated terms (annoyance vs. Anger vs. Rage). Intensity tracking reveals when emotions peak and what triggers those peaks.
Step 4: Track emotional trajectories. Across an interview or diary study, map how emotions shift. A participant might begin an onboarding narrative with excitement, move through confusion and frustration, and end at resignation or mastery. These emotional arcs are findings in themselves.
Combining with Other Methods
Emotion coding pairs naturally with:
- Process coding: capturing what participants were doing when they felt a certain way.
- In vivo coding: when participants use vivid emotional language worth preserving verbatim.
- Values coding: emotions often signal activated values (anger at unfairness signals a value of equity).
- Sentiment analysis: automated sentiment can flag emotionally charged passages for deeper emotion coding.
When to Use Emotion Coding
- Customer experience research: mapping the emotional journey across touchpoints to identify moments of delight and friction.
- Healthcare and patient experience studies: understanding the emotional dimensions of illness, treatment, and care.
- UX research: identifying emotional responses to interfaces, especially frustration, confusion, and satisfaction.
- Brand perception research: understanding the feelings people associate with brands, beyond rational attribute ratings.
- Focus group analysis: tracking emotional energy across a group discussion to identify the topics that generate the strongest reactions.
Common Mistakes
- Over-inferring emotions. If the data doesn't clearly indicate an emotion, explicitly or through strong contextual clues, don't impose one. Code only what the data supports. Uncertain interpretations belong in memos, not the codebook.
- Using too few emotion categories. Reducing everything to "positive" and "negative" misses the point. Frustration, disappointment, and anger are all negative but drive different behaviors. Use specific emotion labels that have actionable implications.
- Ignoring emotional ambivalence. Participants often feel conflicting emotions simultaneously, excited about a new product but anxious about the learning curve. Code both emotions rather than forcing a single classification.
Quali-Fi Support
Quali-Fi's platform combines AI-powered sentiment analysis with human-in-the-loop thematic coding, making it practical to identify emotional patterns across thousands of open-ended responses, focus group transcripts, and discussion board threads. The AI flags emotionally charged passages for researcher review, while the coding interface supports emotion-specific codebooks with intensity tracking.
Understand the emotions driving your customers with Quali-Fi{:.cta-button }
FAQs
How is emotion coding different from sentiment analysis?
Sentiment analysis classifies text as positive, negative, or neutral, it's a broad polarity measure. Emotion coding identifies specific emotions (anger, relief, anxiety, delight) with contextual nuance. Sentiment analysis is automated and works at scale; emotion coding requires human interpretation but produces richer findings. They're complementary, use sentiment analysis to triage and emotion coding to go deep.
What if the participant doesn't express any emotion?
That's data too. Emotional flatness or detachment can be coded as indifference, disengagement, or emotional suppression. The absence of emotion in a context where you'd expect it (describing a service failure with no frustration, for instance) is analytically meaningful.
How many emotion codes should I have?
Most studies work well with 8-15 emotion codes. Fewer than 6 forces overly broad categories that lose specificity. More than 20 makes the codebook unwieldy and reduces intercoder reliability. Start with a core set based on a recognized framework and add domain-specific emotions as they emerge from the data.