How Does Conversation Analysis Work?

The first step of conversation analysis examines the structural form of messages. In linguistics, this is called TCU (Turn-Constructional Unit), which breaks messages down into four levels: Lexical (single words), Phrasal (phrases), Clausal (clauses), and Sentential (full sentences).

People who rapidly fire off short message units lean toward the Quick Ping-Pong type (F), while those who bundle multiple sentences into one long message lean toward the Thoughtful Replier type (S). This is the key factor that determines the “Tempo” axis of your conversation style.

Gap analysis also measures the time difference between receiving a message and sending a reply. The human brain can process roughly 500 words per minute, but actual typing speed is only 125–225 words per minute. By leveraging this gap, we can distinguish whether a delayed response means “thinking carefully” or “not interested.”

In face-to-face communication, non-verbal cues like facial expressions, vocal tone, and gestures account for 93% of communication (Mehrabian's Rule). In text messaging, however, these non-verbal channels disappear. Instead, emojis, text fillers (haha, lol, omg), and punctuation marks (!, ~, ?) take on that non-verbal role.

Research shows that messages containing emojis receive approximately 8 times the weighting on emotional empathy scores. “Congratulations 😊” and “Congratulations” use the same word, but the warmth perceived by the recipient is entirely different. This difference determines the Temperature and Humidity axes.

NLP (Natural Language Processing) techniques also classify text personality. Phrases like “I see” and “That must have been tough” are classified as empathetic (W), while “What's the bottom line?” and “So what do you want me to do about it?” are classified as direct (D).

The third step analyzes the “direction” of conversation. The key metric is the frequency of first-person subject usage (“I”, “me”, “my”). The more often someone shifts the topic to themselves, the higher their Conversation Leader (M) score. Academics call this “Conversational Narcissism.”

Conversely, high use of supportive response phrases (“That's right,” “I see,” “I understand”) and open-ended questions (“How did it go?” “Why do you think so?”) increases the Supportive Listener (Y) score.

The form of questions matters too. Someone who frequently asks “What part did you enjoy?” (open-ended) rather than “Did you enjoy it?” (closed, yes/no) scores higher on Y. Open-ended questions give the other person more room for expression and signal genuine curiosity.

The final step uses the RASA method to comprehensively evaluate scores across all four axes. RASA is a listening framework proposed by communication expert Julian Treasure, consisting of four stages: Receive, Appreciate, Summarize, and Ask.

Receive— Analyzes timestamp gaps between messages. This measures whether sufficient time is taken to fully read and absorb the other person's message, quantifying one's listening posture.

Appreciate— Measures the ratio of emojis, fillers, and exclamations. Higher active engagement signals to the other person: “I'm enjoying this conversation with you.”

Summarize— Extracts the frequency of summary phrases like “so basically,” “in the end,” and “to sum up.” Summarizing someone's story is a confirmation signal that says “I understood you correctly.”

Ask — Tallies the final count of open-ended questions. People who ask more questions are evaluated as having a stronger ability to deepen conversations.