Relationship text-analysis terms should describe evidence, not hide weak claims.

Thread structure and effort terms

Initiation balance: the share of conversation starts attributed to each participant under the report's session-boundary rules. It describes who starts contact in the supplied history; it does not explain schedules, preferences, or agreements outside the thread.

Effort balance: a combined view of observable work such as initiating, asking questions, following up, planning, and carrying repair. A skew can be meaningful, but the metric cannot decide whether the division was voluntary, resented, or appropriate.

Reciprocity: the degree to which one participant's bids, questions, affection, or repair attempts receive a related response. Reciprocity is about exchange structure, not equal word counts in every conversation.

Response rhythm: the distribution and change of reply gaps across the available timeline. A timing shift can support an observation about communication cadence, but it does not prove withdrawal or loss of interest.

Repair, rupture, and conflict terms

Rupture: a sequence where connection or cooperation breaks down through conflict, dismissal, withdrawal, or another observable change. The label identifies a communication event, not who caused the relationship problem.

Repair attempt: a move toward de-escalation, clarification, accountability, reassurance, or renewed contact after tension. Repair success asks whether the exchange becomes workable afterward and whether the same issue immediately returns.

Escalation cluster: a concentrated sequence in which intensity, accusation, contempt-like language, pressure, or rapid hostile turns increase. The evidence window should include what preceded the escalation and how the exchange ended.

Invalidation: wording that dismisses, minimizes, or denies the other participant's stated experience instead of engaging with it. One ambiguous phrase is weaker evidence than a repeated pattern across comparable exchanges.

Warmth, pressure, and emotional-language terms

Positivity reciprocity: how often warm, appreciative, humorous, curious, or reassuring communication is returned. It is a within-thread balance, not a universal relationship-health score.

Coercive pressure: language that corners, threatens, guilts, repeatedly rushes, or makes connection conditional on compliance. Automated analysis can surface the wording and recurrence; it cannot issue a legal or clinical determination.

Emotional tone: the mix and movement of affective language across windows, including warmth, defensiveness, frustration, reassurance, or flatness. Tone classification is vulnerable to sarcasm, dialect, inside jokes, and missing nonverbal cues.

Emotional drift: a sustained change in tone, responsiveness, or reciprocal engagement across comparable periods. Drift names a direction in the record, not the private reason for that direction.

Pattern and trajectory terms

Pattern recurrence: repeated evidence of a similar exchange structure across separate windows. Recurrence gives a finding more support than one isolated example, provided the examples are not duplicates from the export.

Trajectory: the direction of a metric or pattern across time, such as improving repair, lengthening post-conflict gaps, or declining mutual initiation. A trajectory needs comparable coverage and should not be extrapolated into a prediction.

Turning point: a period where several related signals change together and remain changed long enough to warrant review. A turning point can organize the timeline without proving which offline event caused it.

Pattern memory: a post-analysis artifact that connects recurring themes, sequence motifs, and evidence references across the completed thread. It supports report continuity; it is not a memory of facts outside the uploaded conversation.

Evidence and uncertainty terms

Evidence window: the relevant message range and surrounding turns used to support a finding. A useful window preserves enough before-and-after context for the user to inspect the classification.

Metric driver: the observable input that moved a result, such as more one-sided starts, longer post-conflict gaps, fewer acknowledged repairs, or a sustained change in warmth language.

Confidence label: a statement about support in the available record, based on recurrence, distribution, attribution quality, timestamp coverage, and agreement across related signals. It is not a probability that an interpretation is true in real life.

Alternative context: another explanation consistent with the same messages, such as work, illness, travel, a platform change, or missing off-thread communication. Naming alternatives keeps an observation from hardening into mind reading.

Attachment language and report boundaries

Anxious-looking, avoidant-looking, or secure-looking describe message sequences that resemble familiar attachment-related communication patterns. They should remain behavioral adjectives, not diagnoses or fixed identities assigned to a participant.

Grey Mirror is a consumer relationship text-analysis product. Its terms organize the supplied conversation and help a user inspect evidence; they do not provide therapy, legal analysis, professional evaluation, or proof of abuse, intent, compatibility, or future outcomes.

Frequently Asked Questions