Behavioral Metrics · 10 min read
Turn-Taking Imbalance in Text Message Threads
How direction balance is counted, what a lopsided thread does and does not indicate, and why the same ratio can describe an anxious sender or a deliberate one.
The short answer
Turn-taking describes how the floor moves between participants. The underlying counts are simple: messages per participant, consecutive runs from one participant before the other responds, and how those figures change across a thread. Textimony measures this with an analyzer named turn_taking, version 1.3, and keeps every figure linked to the messages behind it. The interpretation is where message analysis usually goes wrong. A thread where one person sent eighty per cent of the messages is a fact; what produced it is not in the file. That ratio is equally consistent with a person sending logistics into silence, an anxious sender who writes in bursts, one participant who prefers phone calls, and a thread where the other side replied on a different platform. Direction balance is a strong signal that something is worth reading and a weak basis for any claim about control. Used as an index it is valuable; used as a conclusion it is an overreach the record cannot carry.
Three counts that get conflated
Message share is the crudest: what proportion of messages each participant sent. It is easy to compute and easy to mislead with, because a person who writes one long message and a person who writes eight short ones have said comparable amounts while producing an eight-to-one ratio. Consecutive-run length is more informative. It counts how many messages one participant sends before the other says anything. A thread of alternating single messages and a thread where one person sends nine messages into silence can have identical message shares and describe completely different interactions. Initiation is the third. Who starts a conversation after a quiet period is a different question from who sends more within one, and the two frequently disagree. A participant can initiate rarely and still dominate the threads they take part in, which is why these counts are reported separately rather than combined into a single dominance figure.
What the ratio cannot tell you
It cannot tell you who was distressed. Repeated unanswered messages are consistent with a person escalating pressure and equally consistent with a person frightened by silence, and the two produce similar counts. Assigning a motive from a ratio is the specific mistake this measurement exists to make visible rather than to commit. It cannot tell you who was being reasonable. Message volume has no relationship to the merits of anything said, and the counts are indifferent to content. A participant who sent four messages across a month may have been the one making unreasonable demands, and a participant who sent four hundred may have been answering them; the ratio is identical in both readings and settles neither. It also cannot survive a partial export. Direction balance computed over a trimmed date range measures the trimming as much as the conversation. Where only one side of a thread is available -- a screenshot set, a single-device export of a cross-platform conversation -- the ratio describes the collection method, not the participants.
Reading imbalance responsibly in family-law review
In co-parenting records the useful question is rarely who sent more. It is whether messages that required an answer got one. A parent sending many messages about exchange logistics into silence, and a parent sending many messages of commentary, produce similar volume asymmetry and mean entirely different things for a parenting-time question. That distinction is only visible by reading the messages, which is why turn-taking is treated as a pointer rather than a finding. The measurement identifies the windows where the shape of the conversation changed; a person then reads those windows and decides whether anything there matters. The metric narrows forty thousand messages to a readable set. It does not replace the reading.
Presenting it without inviting an objection
State the denominator every time. A ratio over an unstated period invites the obvious question of what falls outside it, and a summary that cannot answer that question tends not to survive the asking. Report the window, the message counts on both sides, and the export the counts came from. Keep it descriptive in wording as well as in method. "Participant A sent 812 of 1,003 messages between March and June" is a statement about a file. "Participant A dominated the conversation" is a characterisation, and once a summary starts characterising, the reader is entitled to test the characterisation rather than the count. Under Rule 1006 the underlying records stay available precisely so that test can be run.
Does sending more messages mean someone was controlling?
No. Volume asymmetry is consistent with control, with anxiety, with one participant preferring phone calls, and with a cross-platform conversation captured on only one platform. The count is real; the explanation is not in the file. Any claim about control has to come from the content of the messages and the surrounding circumstances.
What is a one-sided text conversation?
In measurement terms it is a thread with long consecutive runs from one participant and few replies from the other. That shape is worth reading, because it can reflect unanswered logistics, unwanted repeated contact, or a conversation that continued elsewhere. Which one applies is a question for the messages, not the ratio.
How does Textimony measure turn-taking?
The turn_taking analyzer, version 1.3, reports message share per participant, consecutive-run lengths before a reply, and how those change over the thread. The figures stay linked to their source messages so any window can be opened and read.
Why report runs separately from message share?
Because they disagree in the cases that matter. Two threads can have the same overall share while one alternates evenly and the other has one participant sending nine messages into silence. Collapsing both into a single dominance number hides exactly the difference a reviewer needs.
Does this work if I only have screenshots of one side?
Not meaningfully. A partial capture produces a ratio that describes what was collected rather than what was exchanged. Direction balance needs a complete thread over a stated window; anything less measures the collection method.
Can turn-taking imbalance be used in a custody case?
Counts drawn from an authenticated export are ordinary record content and can be summarised. Whether an imbalance is relevant to a parenting question is for counsel and the court. The more durable use is narrowing a long record to the periods worth reading before anyone argues about what they show.
Published by
Textimony. Editorial status: Source-linked informational guide. Updated: 2026-07-30.
Sources
Federal Rule of Evidence 1006: Summaries to Prove Content; Federal Rule of Evidence 403: Excluding Relevant Evidence; California Courts Self-Help: Custody and parenting time