Review Guide · 22 min read
How to review behavioral patterns in text-message evidence.
Textimony can surface candidate messages involving repeated contact, threats, leverage, custody logistics, or stated boundaries. A reviewer must read the full exchange and decide what those messages support.
This page explains how Textimony can surface message candidates for human review and how to discuss volume, timing, language, and sequence without inferring intent, abuse, credibility, or psychological state from a metric alone.
Why false positives matter
A false positive can turn ordinary scheduling, frustration, quoted language, or a consensual exchange into a misleading category. A missed candidate can also matter. Thresholds balance those errors; they do not eliminate them. Textimony keeps automated output labeled as a candidate. A completed run may produce an automatic report before each candidate has a human decision, so that report should be read as analysis output. The review queue and curated rebuild exist so a person can accept, reject, reclassify, or leave an item unresolved after reading context.
Every signal is a candidate, not a verdict
A rule can match words associated with a threat, and a classifier can score a message as similar to examples in a category. Neither component knows the complete relationship history, off-platform events, whether a phrase is quoted, or how a court defines an element. Behavioral counts are counts of candidates under a stated analyzer and run—not counts of proven incidents. Use the count to locate messages, then read the source text, surrounding exchange, participant orientation, date range, and any missing portions before making a human determination.
Sequences can add context without proving a pattern
One message can matter, but sequence often changes how it reads. Repetition after a stated request, a logistical exchange that becomes personal, or an apology followed by renewed disagreement may deserve review as a stretch of conversation rather than as isolated quotations. Textimony can show nearby messages and place candidate counts on a timeline. A token-level span candidate marks words inside one message; it does not automatically identify the beginning and end of a multi-message campaign. The reviewer chooses the context needed to explain the sequence fairly.
Review repeated contact without inventing a legal threshold
Harassment is a legal term whose elements vary by jurisdiction and claim. Message volume, timing, reply direction, and repetition can describe the supplied record, but no universal number of texts or hour of day turns communication into harassment. A responsible review identifies the stated boundary if one exists, counts subsequent messages under a clear date range, includes replies and ordinary logistics, and records gaps in the source. Textimony can surface repeated-contact candidates and daily volume; a person must decide whether the messages are unwanted, permitted, necessary, threatening, or legally significant. Describe the observation first: “The supplied record shows these messages during this period.” Apply a legal label only with the appropriate facts and law.
Review threat and intimidation language in context
Some threat language is explicit; other language depends on history or facts outside the conversation. A model cannot know a private reference simply because the recipient recognizes it. Textimony therefore pairs literal checks and configured classifier candidates with source context rather than claiming to decode hidden meaning. Review the exact words, speaker orientation, target, stated consequence, timing, surrounding requests, quoted material, and relevant off-platform corroboration. Preserve uncertainty when a phrase supports more than one reading. Confirm that the candidate text is a direct message rather than a quotation or forwarded content. Read as much surrounding exchange as the subject requires; there is no universal fixed window. Identify the person, property, action, or consequence referenced without filling in missing facts. Treat reply timing as timing, not proof of fear, agreement, or withdrawal. Record the reviewer’s decision and retain the available source reference.
Review financial leverage language
Financial conversations can include ordinary bills, disputed obligations, bargaining, access to shared property, child expenses, account credentials, or statements tying money to another action. The text can document the words used; it does not by itself establish the obligation, the speaker’s resources, or whether a demand was lawful. Textimony can surface candidates where configured checks associate financial terms with pressure, conditions, or legal-process language. A reviewer should separate the requested action, stated consequence, relevant agreement or order, and later outcome instead of asking the model to decide whether the exchange was cooperative or coercive.
Review custody and parenting-time logistics
Parenting-message records often contain proposed changes, confirmations, cancellations, transportation details, medical information, requests for calls, and disagreements about an order. A candidate can point a reviewer to obstruction or refusal language, but the message record alone may not show the governing schedule or what happened offline. Build a chronology using the stated date, requested change, response, and any later confirmation. Compare it with the relevant agreement or order outside Textimony. Avoid describing a late change as intentional interference, a dense exchange as parental harassment, or a refusal as contempt without the additional facts and legal determination those labels require. Schedule change — When a change was proposed and how the other participant replied. — Whether notice was sufficient or the change violated an agreement. High-volume exchange — Message direction and volume around a transfer or dispute. — Why messages were sent and whether the contact was unwanted. Refusal language — The exact request, response, and surrounding logistics. — Whether an order applied, whether an exception existed, and what occurred offline.
Contradictions and disputed recollections
Two messages can conflict about a date, promise, payment, or prior statement. That observable contradiction may be worth documenting, but it does not establish why the accounts differ. Memory error, ambiguity, changed circumstances, incomplete context, and deliberate deception can produce similar text. Use search and timeline tools to locate the relevant messages, quote both accurately, and state the point of disagreement. Textimony does not run a verified contradiction model that establishes “gaslighting,” and a checksum cannot make a disputed interpretation mathematically certain. A searchable timeline can document inconsistent statements. It cannot diagnose manipulation or prove the speaker’s state of mind.
Changes in tone are observations, not motives
A conversation can move from logistics to accusation, apology, renewed dispute, or silence. Those shifts may help a reviewer choose a useful context window, but they do not form a universal abuse cycle and do not reveal a speaker’s motive. Textimony’s daily counts and candidate lanes can help locate periods with more flagged language. The product does not calculate a clinical volatility curve or certify a threat-apology pattern. Describe the sequence that appears in the record and leave psychological interpretation to qualified, matter-specific review.
Review a stated boundary and what followed
A clear request about contact, topic, channel, location, or timing can provide an anchor for later review. Record the exact wording, exceptions, date, participant, and the messages that followed. Do not silently expand a narrow request into a complete no-contact boundary. Configured checks can surface boundary-related language and repeated-contact candidates, but Textimony does not statefully adjudicate every later message as compliant or noncompliant. A reviewer must determine scope, identify exceptions, account for replies or necessary logistics, and apply any relevant order or law.
Describe change over time without predicting risk
Daily message volume, direction, and candidate counts can rise or fall. A chart can describe the size and date of that change under the frozen case timezone. It cannot identify a precise trigger, assign hostility, or predict a future crisis without further evidence. When a period changes sharply, compare ordinary messages, candidate messages, missing-data warnings, participant orientation, and known external events. Avoid invented cutoffs such as a universal reply speed or density threshold. The useful question is “What changed in this supplied record, and what context explains it?”
Published by
Textimony. Editorial status: Product review guide for observable message language. It does not diagnose a person or define the elements of a legal claim. Updated: 2026-07-23.
Sources
Federal Rule of Evidence 901 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 902 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 1006 — Legal Information Institute, Cornell Law School; Federal Rule of Evidence 106 — Legal Information Institute, Cornell Law School; Guidelines on Mobile Device Forensics — National Institute of Standards and Technology; SWGDE Best Practices for Mobile Device Evidence Collection and Preservation — Scientific Working Group on Digital Evidence