What the data actually shows
Trust researchers Roy Lewicki and Barbara Bunker described trust repair as a multi-step process rather than a single act: acknowledging that a violation occurred, determining its cause, accepting responsibility, and then — crucially — demonstrating over time that the behaviour has changed. In their framework an apology opens the door, but it is the consistent follow-through afterward that does the rebuilding.
Experimental work in this tradition finds that the type of violation matters. Trust harmed by something framed as a competence lapse (an honest error) tends to recover more readily than trust harmed by something framed as an integrity or honesty violation, which people treat as more diagnostic of who someone really is. Once a breach is read as 'this is the kind of person you are,' later good behaviour is discounted more heavily.
The research also points to a sobering asymmetry: trust is generally slower to build than to break, and repair is fragile. Studies in this area find that when an apology and a promise of change are followed by even one further violation, trust often falls back below where it was — the second breach is treated as confirmation. None of this means repair is impossible; it means the evidence describes it as gradual and easily set back rather than fast or guaranteed.
Why this feels different from how it actually is
Repair feels slower than you expect because trust is partly a prediction about the future, and a single breach has already proven that the prediction can be wrong. The injured person is not being stubborn; they are rationally weighting new evidence against a track record that now includes a violation. Words cannot quickly outweigh that, because words were available before the breach too.
It also feels uneven because the two people are running different clocks. The person who caused the harm often experiences the apology as the turning point and wants to move on from there; the person who was harmed experiences the apology as the start of a waiting period in which they watch for the behaviour to actually change. That gap in timelines is one of the most common sources of frustration in trust repair, and it is largely structural rather than a sign that either person is acting in bad faith.
And because integrity violations get discounted so heavily, a sincere effort can feel unfairly received. When trust is rebuilding, the injured person tends to notice and weight any slip more than any success — not out of unfairness, but because they are scanning for whether the original pattern is still there.
An apology opens the door; it is the consistent follow-through afterward that does the rebuilding.
What the research says to do about it
The most consistent finding is that an effective apology and demonstrated change work together, not separately. Research on apologies suggests the components that matter most are a clear acknowledgement of the specific harm, accepting responsibility without deflection or excuse, and a concrete offer to repair or change — followed by actually doing it. The apology signals intent; the subsequent behaviour supplies the evidence.
Consistency over time appears to do the heavy lifting. Because the injured person is updating a prediction about the future, what shifts that prediction is a run of reliable, observable actions rather than a single grand gesture. The research frames trust repair as accumulating small confirmations, which is why patience and repetition tend to matter more than intensity.
Where the breach touched integrity, explicitly naming and addressing the underlying issue — rather than only apologising for the incident — tends to help, because it speaks to the character inference that is doing the damage. Some breaches also benefit from external structure (transparency, agreed checkpoints) that lets the injured person verify change rather than having to take it on faith while faith is exactly what is depleted.
What the research says does not help
A verbal apology on its own, however sincere, is consistently found to be insufficient when it is not followed by changed behaviour. The research is fairly clear that repeated apologies without demonstrated change tend to erode trust further, because they become evidence that the words and the actions are decoupled.
Rushing or pressuring the other person to 'move on' tends to backfire. Since repair depends on the injured party accumulating their own evidence over time, demands to be trusted again immediately ask them to skip the very process that rebuilds trust, and often read as not having grasped the harm.
Over-explaining, minimising, or attaching conditions to the apology ('I'm sorry, but…') reliably weakens it in the research, because it shifts the frame from accepting responsibility toward defending oneself. And treating one good stretch as the finish line is risky: because repair is fragile, a further breach after partial recovery tends to undo much of the progress.
Trust is slower to build than to break, and repair is easily set back — gradual and conditional, not fast or guaranteed.
What this looks like in real life
The apology that felt like the finish line
One partner apologises sincerely and expects the relationship to reset from there. The other hears the same apology as the start of a waiting period — a time to watch whether the behaviour actually changes. Neither is acting in bad faith; they are simply running different clocks, which is one of the most common and most structural sources of frustration in trust repair.
Why the second slip lands so hard
After a breach, an apology, and a promise to change, a single further violation tends to be treated not as a small setback but as confirmation that the change was never real. That is why repair is described as fragile: the injured person is scanning for whether the original pattern is still there, and a repeat breach can drop trust below its post-breach level.
Real numbers in context
This is an area where the honest answer is mostly qualitative rather than numerical — there is no reliable 'percentage of relationships that recover' figure, and you should be wary of anyone who quotes one. What the research offers instead is a direction and a shape: trust is slower to rebuild than to lose, integrity violations repair less readily than competence ones, and repair is set back sharply by repeat breaches.
The broader context worth holding is that conflict and rupture are normal features of close relationships, not signs they are doomed; what tends to distinguish relationships that recover is the presence of repair attempts and follow-through, not the absence of breaches. Treat the timeline as months rather than moments, and treat the absence of a clean, guaranteed answer as the honest state of the evidence.