Behavioral Metrics · 10 min read
Response Time Patterns in Text Message Evidence
What the gap between a message and its reply can support in review, how time zones and platform behaviour distort it, and why latency never establishes intent.
The short answer
Response latency is the elapsed time between a message and the next message from the other participant. It is a direct measurement: subtraction over two timestamps already in the export, not an inference about anybody. Textimony computes it with an analyzer named response_latency, version 1.3, and reports it alongside the messages it came from so a reviewer can open any gap and read what surrounds it. What it supports is narrow and worth stating precisely. A latency distribution can show that replies in one period were typically minutes apart and in another typically days apart. That is a fact about the record. It is not a fact about why. A long gap is equally consistent with sleep, a flight, a dead battery, a deliberate silence, a message read and ignored, and a reply sent on a different platform entirely. The measurement cannot distinguish between those, and any review treating a delay as proof of a state of mind has left the record behind.
What actually gets measured
The calculation needs a definition of "reply" that survives contact with a real thread. Textimony pairs each message with the next message from a different participant, so a run of five consecutive messages from one person produces one latency measured from the last of them, not five overlapping intervals. Without that rule, a person who sends five short messages in a row appears to have five separate response times, and the distribution is dominated by an artifact of typing style. Timestamps are handled as unix epoch milliseconds in UTC, and the time zone is confirmed during case setup rather than guessed from the file. This matters more than it sounds. An export that renders local time without an offset will silently shift every gap that crosses a daylight-saving boundary, and a thread spanning a move between time zones can produce negative intervals if the source is trusted uncritically.
The four things that break it
Platform delivery is the first. A message queued on a device with no signal carries the send time, the receipt time, or both, depending on the platform and the export path. The gap you measure may be the gap between composition and composition, not between reading and replying. Cross-platform conversation is the second. Two people who move between SMS, a messaging app, and a phone call leave a single-platform export that shows a long silence where a conversation actually continued somewhere else. The export is complete about itself and incomplete about the relationship. Partial exports are the third. A thread trimmed to a date range has an artificial first gap, because the message that prompted the first reply is outside the file. Group threads are the fourth: with three or more participants, "the next message from someone else" is not the same as "the reply", and latency in a group thread should be read as ambient responsiveness rather than a directed answer.
Where the measurement is genuinely useful
Latency earns its place as a navigation tool. In an export of forty thousand messages, nobody reads everything. A latency distribution identifies the days where the rhythm of a conversation changed, and those days are where a reviewer should start reading. The metric is not the finding; it is the index that leads to the passage that might be one. It is also useful as a contradiction check. When a written account says contact was constant through a period, and the record shows median reply gaps of several days through the same period, the discrepancy is worth understanding before anyone relies on either. The resolution may well favour the account -- people move to phone calls -- but the question should be asked from the record rather than around it.
How to present it without overstating it
Report the distribution, not a single average. A mean reply time collapses a thread with consistent ten-minute replies and a thread that alternates between instant answers and three-day silences into the same number, and the second thread is the one worth reading. Medians with an explicit spread survive scrutiny; averages invite a question the average cannot answer. Keep every figure attached to its source messages. Under Federal Rule of Evidence 1006 a summary of voluminous records is offered on the basis that the underlying material is available, which is a documentation obligation before it is a presentation format. A latency chart that cannot be resolved back to the specific messages behind each interval is an assertion rather than a summary.
Does a slow reply mean someone was avoiding the conversation?
No. Latency records elapsed time between two messages and nothing else. Sleep, travel, work, a flat battery, a conversation that moved to a phone call, and a deliberate silence all produce the same measurement. Treating a gap as evidence of avoidance substitutes an interpretation for the record, and the record cannot support it either way.
Can response time be used as evidence in court?
Timestamps in an authenticated export are ordinary record content, and a summary computed from them can be presented like any other summary of voluminous material. What matters is that the underlying messages remain available and the method is stated. Whether any particular pattern is relevant or admissible is a question for counsel and the court, not for the software producing the number.
What is a normal text message response time?
There is no general benchmark worth relying on, and a figure quoted from consumer research about typical reply speed says nothing about a specific pair of people. What is usable is an internal comparison: how this thread behaved during one period against how the same thread behaved during another.
How does Textimony calculate response latency?
The response_latency analyzer, version 1.3, pairs each message with the next message from a different participant and records the elapsed time in UTC. Consecutive messages from the same person collapse into a single interval measured from the last one, so typing style does not inflate the distribution.
Do time zones affect the measurement?
Substantially, if they are not handled. Exports that render local time without an offset shift every interval crossing a daylight-saving change, and threads spanning a relocation can produce impossible intervals. Time-zone handling is confirmed during case setup rather than inferred from the file.
Does this work on group threads?
It runs, but read it differently. In a thread with three or more participants the next message from someone else is not necessarily a reply to the previous one, so the figure describes general responsiveness in the thread rather than a directed answer between two people.
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 901: Authenticating or Identifying Evidence; NIST SP 800-101r1: Guidelines on Mobile Device Forensics