How do you apologise properly?
The short answer
Taking responsibility. When researchers tested apologies part by part, an explicit acknowledgement that it was your fault carried the most weight, an offer to repair the damage came second, and asking to be forgiven mattered least. Most apologies are built the other way round, leading with regret and touching responsibility lightly or skipping it. The forms that fail most reliably express regret while withholding responsibility: sorry you feel that way, sorry if I hurt you.
What readers think
When you apologise, do you say plainly that it was your fault, with nothing added after it?
Anonymous. One vote per reader.
Most people asking this have already apologised and watched it fail. The words came out, the other person did not soften, and within a minute the argument had restarted on new ground about whether the apology counted. That sequence is common enough to be worth treating as information rather than as bad luck.
There is a small experimental literature that takes apologies apart into their components and tests which components actually do the work. It is not a large body of evidence, most of it uses hypothetical scenarios, and very little of it was collected on couples, which limits how far it should be pushed. What it does show points consistently in one direction.
The component people put first carries the least weight. The component people work hardest to avoid carries the most.
The parts of an apology are not equally load-bearing
Researchers at Ohio State assembled six components that recur across earlier definitions of an apology: an expression of regret, an explanation of what went wrong, an acknowledgement of responsibility, a declaration of repentance, an offer of repair, and a request for forgiveness. They built apologies containing one, three, or all six, and asked people to rate how adequate each was[1]. Two studies, one with 333 adults recruited online and one with 422 undergraduates, produced the same ranking. More components beat fewer, which is unsurprising. The result worth having is that the components were not interchangeable. Acknowledgement of responsibility, meaning a plain statement that you were wrong and it was your doing, ranked highest. An offer of repair ranked second. The request for forgiveness ranked last, which is to say it is the one you could drop at least cost. An earlier study varying the wording of apologies for a laboratory offence had already pointed the same way[2].
The caveat is usually dropped when this research is repeated. Both Ohio State studies used a workplace scenario about a job applicant who had once filed an incorrect tax return, which is a trust violation between near strangers with a clear factual answer about who did what. A partner apologising for something said at midnight after eleven years together is in a different situation, and nobody has run the equivalent study on them.
Why responsibility is the part that goes missing
If responsibility is what works, the question is why so few apologies contain much of it. A review of the research on offering apologies proposes three barriers: low concern for the person harmed or for the relationship, a perceived threat to the apologiser's self-image, and a belief that apologising will not work anyway[3]. The second shapes most apologies between people who do care about each other. Saying it was my fault requires holding, for the length of a sentence, the idea that you are someone who did that. Most people cannot hold it, so the sentence comes out with something attached that takes it back.
That this is a self-protection problem rather than a vocabulary problem has direct support. In an experiment where participants completed a self-affirmation exercise before responding to someone they had wronged, the affirmed group used more of the core components, remorse, responsibility and repair, and were rated as more sincerely remorseful than controls[4]. Nothing was taught about how to apologise. The threat was reduced and better apologies followed.
This also accounts for the two phrasings everyone recognises. Sorry you feel that way expresses regret about the other person's reaction and locates the problem there rather than in anything the speaker did. Sorry if I hurt you makes the harm hypothetical, which quietly disputes that it happened. Each pairs the weakest component with the strongest one not merely absent but contradicted. They are not failed apologies. They are defences with an apology's opening words.
An apology does not work independently of what it lands in
A well-built apology is not a mechanism that produces forgiveness on contact, and two findings show why.
The first is about intent. When people believed a transgression was deliberate, an apology made forgiveness less likely rather than more; when they believed it was accidental, the apology helped[5]. If your partner has concluded that you knew what you were doing, an apology enters that reading rather than displacing it, and can be heard as an attempt to get out of something. The work that has to happen first is about the belief, not the phrasing.
The second is about the state of the relationship. In a diary study following 60 married or cohabiting couples through everyday offences, apologies predicted forgiveness only among participants highly satisfied with their relationship, and the route ran through perceived sincerity: satisfied partners rated the same apologies as more sincere, and sincerity predicted forgiveness[6]. That inverts the usual assumption. People treat the apology as the thing that repairs the relationship. Here the relationship's condition determines whether the apology can be heard as genuine at all.
The surrounding literature is consistent with that. A meta-analysis places apology among several situational correlates of forgiveness rather than as the dominant one[7], which components land varies with how the recipient thinks about themselves in relation to others[8], and in research on real betrayals within couples, amends and forgiveness developed over time as a two-sided process rather than resolving at the moment of the apology[9].
Why people who apologise badly rarely find out
There is a feedback problem underneath all of this, and it may be the most practically important thing in the literature.
In a set of studies on questionable apologies, participants either received or watched a staged offence followed by a spontaneous apology, a coerced apology, or none. The coerced version was engineered to be obviously hollow: a bystander told the offender they had to apologise, and the offender said sorry. Observers rated it no better than no apology at all. The people who had actually been insulted rated it nearly as good as the spontaneous one[10]. The explanations the researchers found concern the position the target is in. Targets are motivated to be seen well by others and to feel good about themselves, and social scripts constrain them more than they constrain onlookers. Somebody handed a bad apology is standing inside a script that supplies the next line, and the next line is that it is fine.
So acceptance is weak evidence that an apology worked. If you apologise defensively, the person opposite you is under more social pressure than you are, will usually take it, and you will read that as it having landed. The resentment does not appear in the exchange. It appears three weeks later attached to something else, and by then it looks unreasonable rather than delayed.
It depends, on what?
What changes the answer
Several things decide how much of this applies to you. If the same apology has been offered for the same behaviour more than a few times, the apology is no longer the problem: repair is itself a component, and where a specific change was promised and did not happen, further apologies are discredited rather than undersupplied. Severity matters too, since this research skews heavily toward minor and hypothetical offences, and a staged insult tells you little about a serious betrayal. There is also a question the component research cannot answer, because it studies single apologies in isolation. If one person in a household does nearly all of the apologising, something is being settled about whose account of events counts, and improving that person's technique will not touch it. Culture changes the reading as well, since directness, whether the apology happens privately or in front of family, and whether apologising is understood to cost you standing all vary, and most of these samples were Western. Finally, apology behaves differently inside a controlling relationship. Where remorse arrives reliably after each incident and the behaviour recurs on a cycle, the apology is part of the pattern rather than an interruption to it, and none of the above is designed for that.
Where people disagree
Both sides of it
Reasonable people land in different places on this. Here is the strongest version of each case, not a straw man of the one we disagree with.
The case for yes
- The component ranking gives you something specific and testable to change, which is more useful than being told to be sincere. Sincerity is not an instruction; putting the responsibility sentence in and leaving it unqualified is.
- The finding is counterintuitive in a way that survives contact with experience. Most people can recall an apology that contained a great deal of visible distress and no admission, and can recall that it did not work.
- Two independent studies produced the same ordering, and an earlier study varying apology wording pointed the same way, so the direction of the effect is not resting on a single result.
- Treating the barrier as self-image threat rather than stubbornness is supported by the self-affirmation experiment, and it points toward something the apologiser can actually work on.
The case for no
- The component research was conducted on hypothetical workplace transgressions with student and online samples, and generalising a ranking derived from a stranger's tax error to an intimate relationship is a real leap that the studies do not license.
- Optimising the wording risks producing a technically complete apology that is heard as a performance, and the diary research suggests perceived sincerity is what actually predicts forgiveness.
- Where the offence is read as intentional, an apology can make things worse rather than better, so the advice to apologise well is not universally safe advice.
- Focusing on the apology can obscure the more consequential question of what changed afterwards, and a strong apology followed by nothing is worse than a weak one followed by a change.
Evidence
What the research says
- Moderate evidence333 adults recruited online and 422 undergraduates
Across two studies testing apologies built from one, three or six components, acknowledgement of responsibility was rated the most important component and a request for forgiveness the least important, with an offer of repair second.
Lewicki, R. J., Polin, B., & Lount, R. B., Jr., An exploration of the structure of effective apologies (2016)(opens in a new tab)What this does not show
Both studies used a single hypothetical workplace scenario about a job applicant who had filed an incorrect tax return, judged by online panel and undergraduate participants. It tests apologies between near strangers over a factual error, not between partners, and rated adequacy rather than any observed outcome.
- Moderate evidenceReview of published studies
A review of the research on offering apologies identifies three barriers to giving a high-quality one: low concern for the victim or the relationship, perceived threat to the apologiser's self-image, and a belief that the apology will not work.
Schumann, K., The psychology of offering an apology: Understanding the barriers to apologizing and how to overcome them (2018)(opens in a new tab)What this does not show
A narrative review by a single author rather than a systematic synthesis with pooled effects, drawing on a research area the author works in. The three barriers organise existing findings rather than being tested against each other.
- Limited evidenceLaboratory experiments with student and online participants
Participants who completed a self-affirmation exercise before responding to someone they had wronged used more of the core apology components and were rated as more sincerely remorseful than unaffirmed participants.
Schumann, K., An affirmed self and a better apology: The effect of self-affirmation on transgressors' responses to victims (2014)(opens in a new tab)What this does not show
Experimental, with the outcome being the quality of the apology as coded and rated rather than whether the wronged person forgave. Self-affirmation effects are known to vary considerably across contexts and have been inconsistent in replication attempts more broadly.
- Moderate evidenceMultiple experimental studies
When a transgression was believed to be intentional, an apology reduced the likelihood of forgiveness; when it was believed to be unintentional, an apology increased it.
Struthers, C. W., Eaton, J., Santelli, A. G., Uchiyama, M., & Shirvani, N., The effects of attributions of intent and apology on forgiveness: When saying sorry may not help the story (2008)(opens in a new tab)What this does not show
Based on scenario and laboratory studies where intent was manipulated by the experimenters, which is cleaner than real life, where intent is inferred and contested. Says nothing about repeated offences within an ongoing relationship.
- Limited evidence60 married or cohabiting couples
In a daily diary study of couples, partners' apologies predicted forgiveness only among participants high in relationship satisfaction, and the association ran through how sincere the apology was perceived to be.
Schumann, K., Does love mean never having to say you're sorry? Associations between relationship satisfaction, perceived apology sincerity, and forgiveness (2012)(opens in a new tab)What this does not show
A single study of 60 couples over a short diary period, relying on self-report from both sides. Correlational, so the possibility that people who forgive readily also report higher satisfaction is not excluded. Offences reported in daily diaries are mostly minor.
- Moderate evidenceFive studies with student participants
People who received an obviously coerced apology after a staged insult rated it nearly as favourably as a spontaneous apology, while observers of the same exchange rated the coerced apology as no better than receiving none.
Risen, J. L., & Gilovich, T., Target and observer differences in the acceptance of questionable apologies (2007)(opens in a new tab)What this does not show
Staged laboratory offences among strangers, which are mild and time-limited compared with a grievance inside a long relationship. The studies measure immediate ratings, not whether the resentment persisted.
- Strong evidenceSynthesis of a large body of published forgiveness studies
A meta-analytic synthesis of the correlates of forgiveness found apology to be one of several situational factors associated with forgiveness, alongside offence severity and relationship characteristics.
Fehr, R., Gelfand, M. J., & Nag, M., The road to forgiveness: A meta-analytic synthesis of its situational and dispositional correlates (2010)(opens in a new tab)What this does not show
Pooled correlations across heterogeneous designs, most of them cross-sectional and self-reported, so it establishes association rather than causal ordering. Dominated by Western samples and by hypothetical or recalled offences.
How we rate evidence strength
- Strong evidence
- Multiple independent studies, including replications or meta-analyses, point the same way.
- Moderate evidence
- Several studies support this, but samples are limited or findings vary by population.
- Limited evidence
- Early, small, or single-study evidence. Treat this as a reasonable hypothesis, not a settled fact.
- Contested
- Credible researchers disagree, or the evidence points in conflicting directions.
Comparative view
Different perspectives
Traditions and disciplines answer this differently. We describe what each one teaches. We are not telling you which is right, and no tradition speaks with a single voice.
What the Evidence Supports
The defensible claim is that apology components are not equivalent and that responsibility outranks regret in how adequate an apology is judged to be, replicated across two samples in one paper and consistent with an earlier study varying apology wording. The indefensible claim is that this is a formula for repairing a relationship. Almost all of the component research uses hypothetical offences among strangers and measures rated adequacy, not forgiveness and not what happened afterwards. The one diary study of real couples found the association held only for already-satisfied partners, which suggests the apology operates within a relationship rather than on it. The direction of the finding is reasonably solid. The magnitude in any real household is unknown.
Therapist Perspective
Couples therapists commonly report that the apology a client believes they gave and the apology their partner received are different objects, and that much of the work is establishing which one happened. Practitioners tend to distinguish apologising for the effect from apologising for the act, and to treat the first as the more common evasion, because it can be delivered with a great deal of feeling while conceding nothing. A related pattern is the apology offered to end a conversation rather than address it, often detectable by its speed. This reflects widespread clinical practice rather than controlled trials.
Secular Perspective
A secular reading treats an apology as a claim about the future as much as an account of the past. Saying it was my fault concedes that a standard exists, that you failed it, and that you can be held to it next time, which is why it is costly and why it carries weight. On this view a conditional apology fails not because it is impolite but because it concedes nothing and therefore commits to nothing. That also limits the exercise. Intent is cheap to produce and expensive to keep, so nobody is obliged to treat a well-formed apology as settling the matter. What settles it is the next occasion when the same pressure arises and something different happens.
Jewish Teaching on Repentance
Jewish teaching on teshuvah, usually translated as repentance or return, sets out a sequence close to what the component research identifies. In Maimonides' formulation repentance requires abandoning the act, regret, verbal confession, and a resolution not to repeat it. For a wrong against another person, the tradition holds that atonement before God does not take effect until the wronged person has been appeased and any material harm made good, which places responsibility and repair ahead of the internal experience of remorse. The custom of approaching the wronged party up to three times puts the obligation on the offender rather than on the person waiting. This is what the tradition teaches, not a claim about what is true.
Not a settled view
Authorities differ on whether the wronged party is obliged to accept a sincere apology, with a strong strand holding that withholding forgiveness from someone who has genuinely repented is itself a fault, and others treating certain injuries as outside that obligation.
Practical steps
What you can actually do
Put the responsibility sentence in and end it there. I was wrong to say that, and it was my fault, is a complete sentence. Anything appended to it, however reasonable, is read as the point of the exchange.
Separate the explanation from the apology in time. Context is a legitimate component, but delivered in the same breath it arrives as an excuse. If it matters, say it the following day when it is not doing defensive work.
Name the specific thing rather than the general category. Sorry for how I spoke to you in front of your sister is a different act from sorry for everything, because the first proves you know what happened.
Offer a repair that is concrete and within your control, and only offer one you will actually do. This is the second strongest component and the only one that can be checked afterwards.
Do not require an answer. The request for forgiveness ranked last for a reason, and asking for it converts the apology into something the other person now owes you a response to.
Treat quick acceptance as inconclusive. People accept bad apologies under social pressure, so if the same subject keeps returning, the earlier apology probably did not land no matter how it ended.
Worth unlearning
Common misconceptions
That the expression of regret is the core of an apology. It ranked among the weaker components, and on its own it is what most failed apologies consist of.
That saying sorry you feel that way is a clumsy apology. It is not an apology at a component level, because it expresses regret while placing the problem in the other person's reaction.
That a conditional apology is a cautious one. Adding if to the harm disputes that the harm occurred, which contradicts the component doing the most work.
That an apology accepted is an apology that worked. Targets of hollow apologies accept them at far higher rates than onlookers judge them acceptable, so acceptance is a weak signal.
That apologising more will eventually resolve a recurring issue. Where a repair was promised and not delivered, additional apologies carry less weight rather than more.
That the research tells you what will work on your partner. Almost all of it uses hypothetical offences among strangers, and the one diary study on real couples found the effect depended on how satisfied the couple already was.
The short version
Key takeaways
Acknowledging responsibility ranked as the most important component of an apology, and asking for forgiveness ranked as the least.
ResearchAn offer of repair ranked second, which makes it the strongest thing you can add after taking responsibility.
ResearchSorry you feel that way and sorry if I hurt you fail structurally, not tonally: each pairs the weakest component with an explicit denial of the strongest.
By definitionApologies made forgiveness less likely when the harm was believed to be deliberate, so the reading of intent has to be addressed before the wording is.
ResearchIn a diary study of 60 couples, apologies predicted forgiveness only among partners already highly satisfied with the relationship, through perceived sincerity.
ResearchPeople who are hurt accept hollow apologies at much higher rates than observers rate them acceptable, so acceptance is poor feedback on whether it worked.
Research
Worth saying
When to get professional help
If apologies in your relationship follow incidents involving physical force, threats, or being made afraid, that is a cycle rather than a repair process, and a domestic abuse service is the right first contact rather than couples counselling, which is generally advised against where there is fear because it can raise risk. Short of that, couples counselling is worth considering when apologies have stopped registering entirely, when the same one has been given repeatedly for a behaviour that has not changed, or when one person has stopped apologising because it never goes well. Individual therapy is often more useful when you can see that taking responsibility feels intolerable rather than merely difficult, since that reaction usually predates the relationship.
Still wondering
Related questions people ask
There is no good evidence on timing, so this is reasoning from the components rather than a finding. An immediate apology delivered while you are still defending yourself will usually contain the weakest components, because responsibility is hardest to concede when your case is still live. Waiting is only useful if you use the time to arrive at the responsibility sentence rather than to build a better argument.
Community
What other people say
These are readers describing their own relationships, not professionals. Take them as experience, not advice, and please do not name anyone.
No one has shared their experience yet. If you have been here, you would be the first, and probably the most useful thing on this page.
Check our work
Sources
- [1]An exploration of the structure of effective apologies(opens in a new tab)
Lewicki, R. J., Polin, B., & Lount, R. B., Jr. · Negotiation and Conflict Management Research · 2016
- [2]How effective are the things people say to apologize? Effects of the realization of the apology speech act(opens in a new tab)
Scher, S. J., & Darley, J. M. · Journal of Psycholinguistic Research · 1997
- [3]The psychology of offering an apology: Understanding the barriers to apologizing and how to overcome them(opens in a new tab)
Schumann, K. · Current Directions in Psychological Science · 2018
- [4]An affirmed self and a better apology: The effect of self-affirmation on transgressors' responses to victims(opens in a new tab)
Schumann, K. · Journal of Experimental Social Psychology · 2014
- [5]The effects of attributions of intent and apology on forgiveness: When saying sorry may not help the story(opens in a new tab)
Struthers, C. W., Eaton, J., Santelli, A. G., Uchiyama, M., & Shirvani, N. · Journal of Experimental Social Psychology · 2008
- [6]Does love mean never having to say you're sorry? Associations between relationship satisfaction, perceived apology sincerity, and forgiveness(opens in a new tab)
Schumann, K. · Journal of Social and Personal Relationships · 2012
- [7]The road to forgiveness: A meta-analytic synthesis of its situational and dispositional correlates(opens in a new tab)
Fehr, R., Gelfand, M. J., & Nag, M. · Psychological Bulletin · 2010
- [8]When apologies work: How matching apology components to victims' self-construals facilitates forgiveness(opens in a new tab)
Fehr, R., & Gelfand, M. J. · Organizational Behavior and Human Decision Processes · 2010
- [9]In the wake of betrayal: Amends, forgiveness, and the resolution of betrayal(opens in a new tab)
Hannon, P. A., Rusbult, C. E., Finkel, E. J., & Kamashiro, M. · Personal Relationships · 2010
- [10]Target and observer differences in the acceptance of questionable apologies(opens in a new tab)
Risen, J. L., & Gilovich, T. · Journal of Personality and Social Psychology · 2007