Chapter 10: Repair After Difficult Conflict
Part II: Interaction and Context
A difficult conversation can stop while its consequences remain. Voices are quieter. Someone has said, “I’m sorry.” The immediate argument is over, but a hurtful remark, a broken agreement, or the original problem may still need attention. Calling this moment repaired can conceal how much remains unsettled.
Repair is a broad relationship term for efforts to address disconnection, misunderstanding, harm, or the aftermath of conflict. It can involve acknowledging what happened, accepting responsibility, making amends, or changing subsequent conduct. An attempt to repair is an effort toward those ends. Whether the effort addresses the particular harm is a separate question.
This chapter examines those distinctions and the research that informs them. It does not offer a sequence that every couple must complete. Repair also has limits: nobody owes forgiveness, renewed trust, physical affection, or continued contact because another person has made an effort.
An attempt and its outcome
A repair attempt might be a correction during an argument: “That was unfair. Let me say what I mean without insulting you.” It might be an acknowledgment afterward: “You asked me to keep that private, and I told someone.” These fictional statements have different immediate purposes. One addresses the direction of an interaction; the other identifies conduct that has already caused harm. Neither statement tells us what happens next.
An attempt can be evaluated against the problem it claims to address. An apology for raising one’s voice does not answer a question about an unexplained expense. Replacing a damaged possession addresses a tangible loss, but does not by itself establish that the person affected feels safe. A friendly exchange after conflict may coexist with an unresolved disagreement. These are distinctions about what the actions address, rather than judgments about whether warmth or apology has any value.
When someone says, “I tried to repair things,” the effort and its intended outcome need to be specified. Did the person correct a statement, acknowledge an injury, offer restitution, or keep a later commitment? The recipient’s response and the condition of the underlying issue remain relevant. Intention, conduct, and outcome should not be folded into one verdict.
Recovery does not settle the issue
Chapter 9 describes responsiveness, regulation, and recovery. One illustration of how narrowly recovery can be measured comes from Salvatore and colleagues’ study of 73 young heterosexual couples. Observers rated how participants moved from a ten-minute conflict discussion into a four-minute conversation about areas of agreement. That task measured a transition away from conflict. It did not establish that the original issue or any harm had been resolved (Salvatore et al., 2011).
For educational purposes, it is useful to keep emotional recovery and resolution separate. Partners may be ready to eat dinner, go to work, or speak about another subject while still disagreeing about what happened. A less hostile interaction does not erase the concern that preceded it. Conversely, accepting responsibility for a specific action does not require the affected person to feel better immediately.
Keeping an issue visible need not require an immediate return to it. In an ordinary safe interaction, a person might say, “I can stop arguing now, but I still want us to address the missed payment.” That statement identifies an unfinished issue. It does not establish an ideal interval for returning to it, and it does not apply a return obligation to someone leaving an unsafe situation.
Accountability identifies the conduct
Accountability concerns what a person did and what remains theirs to address. It should not require partners to divide responsibility equally. If one person disclosed private information, that person’s responsibility for the disclosure is not halved because the other person became angry. Each person can be accountable for their own conduct without a requirement for matching admissions.
Acknowledgment makes the object of repair clearer. “I can see you are upset” recognizes a reported state. “I told someone information you had asked me to keep private” identifies an action. “You said that left you feeling exposed” attends to its reported impact. Those statements answer different questions, and none requires a claim to know the other person’s entire inner experience.
An explanation can supply context while leaving responsibility intact. “I was distracted” may describe circumstances around a missed commitment. It does not turn the missed commitment into the other person’s fault. Likewise, stress, an attachment-related concern, or a familiar conflict response does not excuse an insult, coercion, or violence. Understanding a behavior’s possible function cannot decide whether the behavior was acceptable.
Accountability also does not require agreement with an inaccurate accusation. A person can acknowledge interrupting without accepting that every disagreement was deliberate humiliation. Where facts or intentions remain uncertain, the distinction can be stated plainly: “I interrupted you several times. I did not intend to embarrass you, and I hear that you felt embarrassed.” The acknowledgment should not be used to dismiss the reported impact, while the report of impact should not automatically be treated as proof of intent.
What apology research contributes
Lewis and colleagues’ review identifies several recurring components across descriptions of apologies: acknowledgment of wrongdoing, acceptance of responsibility, remorse, an offer of compensation, and a stated intention not to repeat the conduct. The authors also describe substantial variation in how apologies are defined. These components are a way to examine what an apology communicates, rather than a validated sequence of necessary and sufficient steps (Lewis et al., 2015).
The review also has important limits. Much of the research used college samples and hypothetical situations or retrospective reports. Designs were predominantly cross-sectional, meaning variables were assessed at one point in time rather than followed across repeated assessments. Different kinds of close relationships were often grouped together. The review therefore does not establish a single best apology for every couple or show that including particular words prevents repeated harm (Lewis et al., 2015).
Consider the fictional contrast between “I’m sorry you took it that way” and “I shared the information after agreeing to keep it private. I’m sorry I did that.” The second statement identifies the speaker’s conduct and accepts responsibility for it. The comparison illustrates an acknowledgment, without establishing its effect on the recipient or the speaker’s later behavior.
An apology can also include an offer to address consequences: correcting information given to someone else, replacing something damaged, or completing an obligation that remains unmet. Whether a proposed action is relevant depends on the harm. A gift does not answer every kind of injury, and an offer does not give its maker authority to decide that the matter is closed. The affected person can disagree with the proposal or decline involvement.
Making amends is broader than saying sorry
Research on amends looks beyond whether an apology was uttered. In one study by Hannon and colleagues, 78 undergraduates in dating relationships recorded hurtful incidents over two weeks. Reported amends was associated with greater perceived resolution of an incident even after the analysis accounted for reported forgiveness. This was an association in diary reports, not a randomized test showing that an amends technique caused resolution (Hannon et al., 2010).
That work examined ongoing relationships, and the diary incidents were comparatively mild. Its findings do not establish that a particular response prevents repeated harm or makes reconciliation appropriate after severe betrayal. The broader research program also distinguished amends from an explicit verbal apology. For this chapter, its contribution is bounded: studying what happens after harm requires attention to more than the presence of the words “I’m sorry” (Hannon et al., 2010).
The practical distinction is between describing regret and addressing what regret concerns. If a partner missed an agreed responsibility, an acknowledgment can identify the missed action. An offer to complete or redistribute the remaining task concerns its consequences. A later commitment concerns future conduct. These actions are related, but treating one as proof of the others makes accountability harder to describe accurately.
Responsiveness includes hearing the response
Responsiveness concerns whether someone experiences a partner as understanding, validating, caring for, or responding to their needs and concerns. Chapter 9 develops that concept. In repair, an educational illustration is allowing the affected person to explain what remains unaddressed instead of declaring the attempt complete on their behalf. Responsiveness does not require agreement with every interpretation or compliance with every request.
Suppose the person who disclosed private information says, “I thought apologizing was what you needed.” The other person replies, “I appreciate the apology. I’m still worried about who knows.” A response that attends to this concern would address the disclosure and its consequences. Insisting that the apology should already have removed the worry asks the recipient to supply an outcome the speaker cannot establish merely by intending it.
The recipient is also entitled to correct an inaccurate account of their experience. “It wasn’t only embarrassing. I had a reason to keep it private.” No one has to infer which need is operating when the person is willing and able to explain it. Equally, an invitation to explain is not a demand for further disclosure, continued discussion, or reassurance for the person who caused the harm.
A promise and follow-through answer different questions
An intention to change describes proposed future conduct. Follow-through concerns what actually occurs. This distinction matters even when the promise is sincere. “I will not share that information again” is a commitment. Whether another disclosure occurs is a subsequent fact. Neither the forcefulness of the promise nor the elegance of the apology can substitute for that fact.
Giacobbi and Lalot’s systematic review of trust repair identified 13 articles, mostly qualitative and focused on fidelity-related trust. Remorse, accountability, and dependable conduct appeared in the review’s thematic treatment. Small, varied samples and a search limited to English-language publications in two databases constrain its conclusions. The themes are not a pooled test of a repair formula, and findings about infidelity should not be generalized automatically to every disagreement (Giacobbi & Lalot, 2025).
An educational example can make a commitment more specific without claiming that specificity guarantees change. After a missed payment, “I will handle it” leaves the proposed action unclear. “I will make the payment tomorrow and tell you whether it went through” identifies a task and a way to report its completion. The example specifies what completion would mean.
Different responsibilities
The person responsible for the action remains responsible for carrying it out. A commitment does not automatically turn the affected person into its manager. They may choose to discuss an arrangement, ask questions, or decide that a proposal is insufficient. They are not obliged to supervise another adult’s conduct in order to make accountability possible.
Follow-through should also remain tied to the relevant conduct. Keeping an unrelated promise does not establish that a privacy agreement was honored. Providing affection does not show that an unpaid bill was paid. If an agreement is missed again, the new event needs its own accurate account. The earlier apology remains part of the history, but cannot certify what happened afterward.
Returning to the substantive issue
Repairing the way a disagreement unfolded can leave the disagreement itself unfinished. An apology for ridicule during a discussion about household work does not decide how that work should be allocated. Both questions can remain legitimate: what happened during the conversation, and what arrangement the partners are trying to reach.
In a fictional ordinary disagreement, one person might say, “Calling you lazy was wrong. I still need us to discuss how the work is divided.” The other might acknowledge the apology while disputing the proposed division. The example permits accountability and continuing disagreement to coexist. It does not make the original complaint automatically correct, or require an equal apology before the issue can be discussed.
A possible return arrangement can identify the issue and a time the people choose to reconsider it. Its purpose is to make the unfinished conversation explicit. The research reviewed here does not establish one required timetable or guarantee that returning will produce agreement. A return arrangement is also different from a demand that someone remain available for interrogation, surrender a boundary, or resume contact after a protective exit.
Forgiveness is not evidence of future safety
Treating forgiveness as proof that repair is complete conflates different questions. Forgiveness, renewed trust, continued contact, and reconciliation are distinguishable decisions or experiences. Accepting an apology need not settle all of them, and choosing any one does not establish that harmful conduct has stopped.
Evidence about later conduct requires a separate caution. In McNulty’s study of 72 newlywed couples, a greater reported tendency to express forgiveness at baseline was associated with partner-reported psychological and physical aggression that remained relatively stable across four years; lower forgiveness tendencies were associated with declines. Forgiveness was assessed with hypothetical scenarios, aggression was reported, and only 37 couples participated in the final analyzed wave. These observational findings do not show that forgiveness causes abuse or that withholding forgiveness prevents it (McNulty, 2011).
The conduct remains the responsibility of the person engaging in it. This study is not an instruction to punish a partner, become hostile, or withhold forgiveness strategically. Nor should the association between amends and perceived resolution in Hannon and colleagues’ study become an instruction to forgive (Hannon et al., 2010). Different measures and outcomes cannot be assembled into a rule about what a harmed person must do.
A person can say, “I understand that you are apologizing, and I am not ready to resume contact.” That response does not invalidate the acknowledgment or make its recipient responsible for completing the other person’s repair effort. Physical affection and sexual activity likewise remain matters of consent, not demonstrations that an apology has been accepted.
When joint repair is the wrong starting point
Fear, threats, coercive control, stalking, sexual coercion, violence, or fear of retaliation change the appropriate response. These situations should not be treated primarily as a mutual failure to repair. An apology does not establish safety, and shared-pattern language must not distribute responsibility for one person’s harmful acts across both partners.
If someone is afraid of their partner or concerned about retaliation, individualized support from a qualified professional or a specialized domestic violence service may be more appropriate than a joint conversation. Immediate danger calls for emergency assistance. Leaving an unsafe interaction, limiting contact, seeking support, or declining intimacy does not require the other person’s permission. Chapter 53 addresses safety and boundaries in greater depth.
What the evidence leaves open
The reviewed sources study different parts of the aftermath of conflict: a transition into another task, apology concepts, reported amends, perceived incident resolution, trust rebuilding, and later reported aggression. They do not evaluate one unified process with one agreed endpoint. Their population and measurement limits matter when translating findings into ordinary language.
The educational examples in this chapter show distinctions a reader can examine. They do not demonstrate the effectiveness of particular wording or predict what another person will do. Evidence that people report resolution of an incident cannot certify durable change, and a useful conceptual distinction should not be advertised as a validated intervention.
Repair also differs from prevention. Addressing an insult already spoken does not replace responsibility for future speech. A plan for future conduct does not remove the need to acknowledge an injury that already occurred. Repair and prevention concern different parts of the situation; neither makes the other unnecessary.
Chapter 11 introduces the full Break the Loop framework. The external research discussed here does not validate that framework, and repair is not a named BTL sequence. Chapters 41–43 return to aftermath and repair in application. The distinction to carry forward is concrete: an effort, the response to that effort, the original issue, and subsequent conduct each require attention on their own terms.
Back to the Field Guide table of contents
Previous Chapter 9: Responsiveness, Regulation, and Recovery