What the data actually shows
Research by Roy Lewicki, Beth Polin and Robert Lount Jr. examined what makes apologies effective by breaking them into components — including expressing regret, explaining what went wrong, acknowledging responsibility, declaring repentance, offering to repair the harm, and requesting forgiveness. Their work found that not all parts carry equal weight: acknowledgment of responsibility stood out as the most important element, and an offer of repair was the next most influential.
The flip side is the non-apology. Statements that shift blame, minimise ('if anyone was offended'), or center the speaker's discomfort rather than the harm tend not to function as apologies at all — and can read as evasions that damage trust further. The form 'I'm sorry you feel that way' is the textbook example: it expresses no responsibility, so it does little of the work an apology is supposed to do.
More broadly, apologies are associated with greater forgiveness and smoother relationship and trust repair, but the link runs through perceived sincerity and through what follows. An apology contradicted by repeated behaviour loses its force; over-apologising for small things, or apologising reflexively, dilutes the signal so that a serious apology later carries less weight.
Why this feels different from how it actually is
Apologising feels harder than it 'should' because the most effective ingredient — taking responsibility — is exactly the part that threatens the apologiser's self-image. It is psychologically easier to explain, justify, or half-apologise than to plainly own the harm, which is why so many real-world apologies drift toward the defensive versions the research shows work least well.
Receiving an apology can feel confusing for the opposite reason: a fluent, well-delivered apology can sound convincing while lacking the substance that predicts real repair. People sometimes accept a polished non-apology in the moment and then find the hurt unresolved, because the words requested forgiveness without acknowledging responsibility or offering to make things right.
And both sides tend to overweight the apology itself and underweight what comes after. Saying sorry feels like the resolution, so it's tempting to treat it as the end of the matter. But the evidence points the other way — the apology opens the door, and changed behaviour over time is what actually rebuilds trust. Without that follow-through, even a sincere apology can feel, in hindsight, like it fixed nothing.
An apology is not a magic reset button; it is a specific act whose effect depends on what's actually in it.
What the research says to do about it
Lead with the two highest-value components: clearly acknowledge responsibility for the specific harm, and offer concretely to repair it. These consistently matter most in the research, and they are also the ones most often skipped. Naming what you did, without 'but' or explanation creeping in to soften it, does more than any other single element.
Make sincerity and timing real, not performed. An apology read as genuine — and offered reasonably promptly rather than dragged out or extracted under pressure — tends to land better. Where an explanation is needed, frame it as information about what happened, not as a justification that quietly relocates the blame.
Then let behaviour carry the rest. Because trust repair runs through what follows the apology, the most reliable thing you can do is change the behaviour that caused the harm and let consistency over time demonstrate it. An apology backed by a different pattern of action repairs trust; one followed by the same behaviour erodes it.
What the research says does not help
Non-apologies actively hurt. 'I'm sorry you feel that way', 'mistakes were made', and 'if anyone was offended' express no responsibility, so they do little of an apology's work and often read as evasions that damage trust further. A defensive apology can leave the relationship worse off than silence.
Adding 'but' and an explanation usually undoes the apology. The moment a justification follows the acknowledgment, it tends to relocate responsibility and signal that you don't fully own the harm — quietly cancelling the most important component the research identifies.
Over-apologising and reflexive apologies dilute the currency. Saying sorry constantly, or for things that don't warrant it, weakens the signal so a serious apology later carries less weight. And an apology with no behaviour change behind it fades quickly — repeating the harm tells the other person the words meant little, regardless of how well they were phrased.
The apology opens the door; changed behaviour over time is what actually rebuilds trust.
What this looks like in real life
The polished non-apology
A fluent, well-delivered 'I'm sorry you feel that way' can sound convincing in the moment and get accepted — then the hurt stays unresolved. The words requested forgiveness without acknowledging responsibility or offering to make things right, so they did almost none of an apology's actual work, and can read as an evasion that damages trust further.
'Sorry, but...'
The apology starts well — a clear acknowledgment — and then a justification follows. The moment the 'but' arrives, it tends to relocate responsibility and signal you don't fully own the harm, quietly cancelling the most important component the research identifies. Naming what you did without the explanation creeping in to soften it does more than any other single element.
A sincere apology, then the same behaviour
Someone owns the harm, offers to make it right, and means it — and then repeats the behaviour. Because trust repair runs through what follows, the words lose their force fast. An apology backed by a different pattern of action rebuilds trust; one contradicted by the same behaviour repeating tells the other person the words meant little, however well phrased.
Real numbers in context
This research is more about structure than headline statistics. The most useful finding from Lewicki, Polin and Lount's component analysis is a ranking, not a percentage: among the parts of an apology, acknowledgment of responsibility was the most important element and an offer of repair the next, with requesting forgiveness among the least influential. Treat that as the shape of an effective apology rather than a precise formula that guarantees forgiveness.
The other reliable pattern is conditional: apologies aid forgiveness and trust repair, but the effect depends on sincerity, timing, and follow-through. Where harm is serious or trust was broken deliberately, words alone do less and demonstrated behaviour change over time does more. The honest summary is that a real apology meaningfully helps, a hollow one can backfire, and no apology survives being contradicted by the same behaviour repeating.
Lewicki, Polin and Lount broke apologies into components and found they don't carry equal weight. This is a ranking of influence, not a scored formula — and the two heavy lifters are also the ones most often skipped.
| Component | Relative influence |
|---|---|
| Acknowledging responsibility | Most important |
| Offering to repair the harm | Next most influential |
| Expressing regret / explaining / declaring repentance | Help less |
| Requesting forgiveness | Among the least influential |