Justin Kruger and colleagues ran a study in which people wrote messages meant to read as sarcastic or serious and then predicted how well the tone would come across. Senders were confident. Receivers were considerably worse at reading tone than senders expected — because when you write a message you hear it in your own voice, complete with the inflection you intended, and that private soundtrack does not travel with it. This is a small problem on 'sounds good.' It is an enormous one on 'fine, do what you want.'
Text also strips out everything that normally ends an argument early. No face to notice softening in, no tone to hear the wobble in, no way to interrupt yourself halfway through a sentence. It adds two things instead: delay, which both of you fill with the least generous available interpretation, and a transcript, which means anything said badly can be re-read at two in the morning and quoted back for months. Very few people argue well in a medium with a permanent record and no eyebrows.
That does not make the rule 'never text about anything real.' If you live apart, work opposite hours, or sit in different time zones, text is often the only channel there is, and refusing to use it just means the conversation never happens. The workable rule is about escalation rather than topic. Text is fine for raising something and fine for repairing something. It becomes unreliable the moment either of you starts getting worked up — and the tell for that is in your own body, not on the screen.
Agree on a switch signal: one phrase either of you can send that means this needs voices, not thumbs. Something plain works best — 'call me when you can, nothing's wrong.' Agree that sending it is never an escalation and never has to be justified. Then pick one thing you will simply never do over text, whatever the circumstances, and say it out loud so it is a shared rule rather than a private hope.
