24 messages · Apr 2 – Jul 28, 2026 · Alex & Jordan
This is the headline balance-model metric. It compares positive and negative affect markers without applying a universal cutoff or converting the result into an outcome estimate.
We split the conversation into equal message windows.
was 5m
was 2.4%
was 40%
was 33.3%
Their median reply moved from 5 to 53 minutes.
This compares one person's recent timing with that same person's established baseline. It is not evidence of declining interest and may reflect expectations, attachment style, schedules, sleep, or conflict context.Uncertainty: Timestamps do not reveal availability, notification settings, urgency, or attachment style.Based on peer-reviewed research · Response-time expectations in mobile messaging between romantic partnersTheir median reply moved from 5 to 47 minutes.
This compares one person's recent timing with that same person's established baseline. It is not evidence of declining interest and may reflect expectations, attachment style, schedules, sleep, or conflict context.Uncertainty: Timestamps do not reveal availability, notification settings, urgency, or attachment style.Based on peer-reviewed research · Response-time expectations in mobile messaging between romantic partnersFunction-word LSM moved from 0.876 to 0.458.
LSM compares each person's distribution across nine function-word categories. Decline is the theoretically meaningful direction; the absolute level is not a relationship grade.Uncertainty: Short samples and topic changes can move function-word distributions.Based on peer-reviewed research · Language Style Matching Predicts Relationship Initiation and StabilityWarmth, humor, affection, appreciation, and interest moved from 83.3% to 50% of messages.
Everyday texting is closer to events-of-the-day conversation than to a laboratory conflict task, so this signal is prioritized over raw conflict volume.Uncertainty: Text cannot capture warmth expressed offline and humor is especially context-sensitive.Based on peer-reviewed research · The Timing of Divorce: Predicting When a Couple Will Divorce Over a 14-Year PeriodRepeated request/topic sequences followed by short, deflecting, delayed, or absent replies moved from 40% to 0%.
This is a sequence-level proxy, not a trait. Withdrawal can be adaptive in some structural contexts, so the meaning is explicitly context-dependent.Uncertainty: The export cannot show unrecorded calls, in-person replies, or the socioeconomic context that changes interpretation.Based on peer-reviewed research · A meta-analytic review of the demand/withdraw pattern of interaction and its associations with individual, relational, and communicative outcomesThe between-partner initiation gap moved from 33.3 to 0 points.
This describes similarity in who starts a new thread after a four-hour gap. It does not mean the less frequent initiator cares less.Uncertainty: Thread starts are shaped by routines and the export may omit other channels.Based on peer-reviewed research · Perceived similarity in texting behavior and relationship satisfactionThe between-partner we/I gap moved from 16.7 to 0 points.
We-talk is not monotonically good. The engine reports each person's I/we/you profile and emphasizes asymmetry rather than awarding points for more 'we.'Uncertainty: Pronoun use varies by topic, language, and conversational role.Based on peer-reviewed research · Meta-analytic evidence that we-talk predicts relationship and personal functioning in romantic couplesThe proxy index moved from 47.2 to 40.6.
The index combines partner-directed questions, relative elaboration, topic follow-up, and emoji use in reply pairs. It does not measure perceived responsiveness directly.Uncertainty: Perceived responsiveness is subjective and cannot be recovered from message features alone.Based on peer-reviewed research · Intimacy as an interpersonal process: The importance of self-disclosure, partner disclosure, and perceived partner responsiveness in interpersonal exchangesPositivity, assurances, shared tasks, and social-network references averaged 16.7% earlier and 16.7% recently.
The four maintenance strategies are retained separately in the metrics and summarized here without prescribing a target frequency.Uncertainty: Maintenance can happen through calls, actions, and in-person interaction that are absent from this text export.Based on peer-reviewed research · Maintenance strategies in the text messages of romantic partnersStonewalling proxies moved from 8.3% to 8.3%.
This proxy uses explicit abandonment, terse deflection, and thread timing. Withdrawal can be adaptive or harmful depending on socioeconomic and situational context.Uncertainty: A lexical marker cannot establish the interactional meaning or a person's character.Based on peer-reviewed research · Why Marriages Succeed or FailUse this as written, or make it sound like you.
Unlock every chart, timeline, pattern, reflection, and question for this analysis.
Payments not configured