Deciding whether to divorce: the stage before anyone has decided
Every other page on this site assumes the decision is made. This one does not, and it is not going to push you in either direction. Being undecided is not weakness or a failure to face facts. It is the normal state, it is where most people considering this actually are, and there is real research on it.
Are you safe? If you are afraid of your partner, if you change what you do to manage their reaction, if you are being monitored, isolated, or controlled financially, then this is not a decision to weigh and treating it as one is part of the harm. Abuse is not a relationship problem that better communication fixes, and couples counseling is actively contraindicated where there is coercive control, because it hands a controlling person a new tool. Go here instead. Nothing else on this page is written for that situation.
Ambivalence Is the Normal State
Real survey research on people considering divorce — most of it from Alan Hawkins (Brigham Young University) and William Doherty (University of Minnesota) working together — finds that most people who are thinking about divorce are not settled on it. Divorce ideation typically moves slowly, fluctuates over time, and a meaningful minority of even the leaning-out partner still holds open some possibility of saving the marriage. ‘Mixed-agenda’ couples, where one partner wants out and one wants to stay, are common enough in practice that a whole clinical model (discernment counseling) was built around them.
About 1 in 4 individuals already in the divorce process still believed their marriage could be saved.
A national sample of parents who had filed for divorce was asked about their reasons for divorcing and their openness to reconciliation. Roughly one-quarter indicated some belief their marriage could still be saved, and the study found real differences by gender and by stated reason for divorce in how open people were to reconciliation.
A separate national study of divorcing parents found meaningful interest in reconciliation even after a divorce petition had been filed.
Doherty, Willoughby, and Peterson surveyed divorcing parents in Hennepin County, Minnesota and found a subset expressing genuine interest in reconciliation services, which is part of the empirical basis for offering ‘second-look’ options such as discernment counseling inside the divorce process itself.
Divorce ideation (‘thinking about divorce’) is often unstable over time rather than a fixed, settled state.
In a national sample of married individuals who said they had thought about divorce, researchers tracked how that thinking changed. Ideation moved up and down over the following months for a substantial share of participants — consistent with ambivalence rather than a linear march toward a decision.
Clinicians who specialize in this population describe ambivalence as the default emotional state of people in troubled marriages, not the exception.
Reported consistently, not settled
Steven Harris, associate director of the University of Minnesota’s Minnesota Couples on the Brink Project, writes that ‘deciding to divorce does not happen overnight, and for most couples the process is fraught with ambivalence,’ and that in only a small fraction of couples seen at the clinic is the decision moving ‘at highway speeds.’ This is a clinical/practice observation from a research center, not itself a controlled study.
Deciding to divorce does not happen overnight, and for most couples the process is fraught with ambivalence.
What to actually do
- Do not treat ‘I’ve been thinking about divorce’ as equivalent to ‘I have decided to divorce’ — research on divorce ideation shows the two are frequently far apart.
- Expect ambivalence to fluctuate week to week; a bad week is not necessarily new information about the relationship.
- If one partner is leaning out and doubts couples therapy can help, standard couples counseling is often the wrong first step — see the discernment counseling topic.
When the Answer Isn’t a Decision At All — Abuse and Coercive Control
Where there is abuse or coercive control, ‘should I stay or go’ is the wrong frame, and standard couples counseling is not simply unhelpful but actively risky. The National Domestic Violence Hotline states this directly and explains the mechanism: couples counseling presumes a level, two-way relationship that abuse structurally rules out, and it can be used by an abusive partner against the person being abused.
The National Domestic Violence Hotline explicitly does not recommend couples counseling for anyone in an abusive relationship, framing abuse as categorically different from an ordinary relationship problem.
The Hotline states abuse is not a relationship problem to be jointly worked on, because doing so requires a level playing field that an abusive dynamic, by definition, does not have.
We at The Hotline do not encourage anyone in an abusive relationship to seek counseling with their partner. Abuse is not a relationship problem.
The National Domestic Violence Hotline. Should I Go To Couples Therapy With My Abusive Partner?
The Hotline identifies specific mechanisms by which joint counseling can increase risk to the abused partner.
Therapy can make a person feel vulnerable, and if an abuser feels embarrassed or angered by what is said in session, they may retaliate afterward to regain a sense of control. The Hotline also notes that the abused partner often cannot accurately rate ‘how the relationship is doing’ because the abuse itself distorts that judgment, and that a facilitator unaware of the abuse can end up believing the abuser’s framing of the couple’s problems.
The National Domestic Violence Hotline. Should I Go To Couples Therapy With My Abusive Partner?
The authoritative domestic-violence source used here is explicit that couples counseling is not recommended in this situation, and explains specific mechanisms of risk — retaliation after sessions, an abuser controlling the narrative with the therapist, and the abused partner’s own judgment being distorted by the abuse itself.
The National Domestic Violence Hotline. Should I Go To Couples Therapy With My Abusive Partner?
What to actually do
- If there is a pattern of coercive control, physical violence, or credible fear of retaliation, route away from any ‘should I stay or should I go’ framing and toward safety planning and individual, trauma-informed support instead.
- The National Domestic Violence Hotline (thehotline.org, 1-800-799-7233) is a safe first stop for anyone unsure whether what they’re experiencing is abuse.
- Do not recommend or link to standard discernment-counseling or couples-therapy resources for a reader describing coercive control — flag this distinction clearly wherever those resources are presented on the site.
Eight questions, and what each one is actually for
These are clinical framings rather than research findings, and we would rather say so than dress them up. They are not a quiz and they do not add up to an answer. Each one is here because it tends to produce a more honest sentence than “should I get divorced” does.
The most useful of them, because it removes the hope that something might change on its own without asking you to decide anything yet. A no here is not a decision to leave. It is information about what you are actually waiting for.
If you cannot name it, that is worth knowing. If you can name it, the next question is whether you have ever said it out loud, plainly, to the person who would have to change it.
A surprising number of people have raised something in a fight and never once outside one. Those are not the same thing and the second has never been tried.
Sometimes the decision was made some time ago by one or both of you, and what is left is the admitting.
Money, the children, being alone, what people will think, being wrong. They need different answers, and several of them are on this site with actual numbers attached.
If the answer to either is someone other than you, that is worth sitting with.
The oldest question in the list and still the one that produces the most honest answer, because most people are far clearer about other people.
Uncomfortable and clarifying. Children learn what a relationship is from the one they live in.
Discernment Counseling
Discernment counseling is a real, manualized, short-term clinical intervention developed by William Doherty at the University of Minnesota specifically for ‘mixed-agenda’ couples — one partner leaning out and doubtful that therapy will help, the other leaning in and wanting to work on the marriage. It is explicitly not couples therapy and does not try to resolve the marriage’s problems in the room.
Discernment counseling is a distinct, structured protocol, not a variant of couples therapy.
It is designed for couples where the partners disagree about whether to work on the marriage at all. Rather than trying to improve communication or solve problems together (the goal of ordinary couples therapy), it aims to help each partner get clarity and confidence about a direction, while explicitly accepting — rather than trying to resolve — the leaning-out partner’s ambivalence.
The structure is short: an initial 2-hour session, with a maximum of five sessions total, ending in one of three defined paths.
Per the official discernment counseling training site (run by the model’s developers), each partner decides separately after the first session whether to continue, up to five sessions maximum. The three possible outcomes are: Path One, status quo (stay married, no further action for now); Path Two, proceed to separation/divorce; Path Three, a six-month commitment to full couples therapy with divorce off the table during that period.
The model has an accompanying research and outreach program, the Minnesota Couples on the Brink Project, with a substantial peer-reviewed publication record.
The project (University of Minnesota, directed by Doherty and Steven Harris) has generated numerous peer-reviewed studies on divorce decision-making, discernment counseling outcomes, and related topics, listed on its own published-research page.
Minnesota Couples on the Brink Project, Published Research and Outreach.
Outcome research on discernment counseling exists but is still limited, mostly qualitative or descriptive rather than randomized-controlled.
Reported consistently, not settled
One published study examined the experience of individuals who went through discernment counseling and ultimately divorced, focusing on post-divorce communication and co-parenting outcomes, rather than a randomized trial against standard couples therapy.
What to actually do
- Discernment counseling is appropriate specifically for mixed-agenda couples (one leaning in, one leaning out) — not for couples who both want to fix the relationship (use ordinary couples therapy) and not where there is abuse or coercive control (see that topic).
- It is short by design: expect a single 2-hour session as the only commitment, with up to four more only if both partners choose to continue.
- It ends in a decision about direction, not a resolved marriage — the six-month reconciliation path (Path Three) is itself followed by full couples therapy, not therapy done within discernment counseling itself.
If one of you wants out and the other wants to stay, ordinary couples therapy is the wrong tool, because it is built for two people trying to fix something. Discernment counseling was designed for exactly this mismatch, it is short and bounded rather than open-ended, and its stated goal is clarity and confidence about a direction rather than saving the marriage. It is a real, manualised intervention out of the University of Minnesota, and the details and sources are above.
What the Research Actually Says About Predicting Divorce
John Gottman’s ‘Four Horsemen’ research (criticism, contempt, defensiveness, stonewalling) is the most famous body of work on predicting divorce, and the Gottman Institute continues to describe contempt specifically as the single strongest predictor identified in their observational research. But the widely repeated ‘90% (or 94%) prediction accuracy’ claim has a documented, serious methodological critique: independent commentators have pointed out that Gottman’s early published ‘predictions’ were generated by fitting a statistical model after the couples’ actual outcomes were already known, not by predicting forward on an unseen sample — the step required to call it a genuine prediction.
The Gottman Institute identifies four specific communication behaviors — criticism, contempt, defensiveness, and stonewalling — as patterns their research associates with relationship breakdown, with contempt singled out as the strongest single predictor.
Reported consistently, not settled
Criticism is defined as an attack on a partner’s character rather than a specific complaint; contempt (mockery, sarcasm, eye-rolling, treating a partner as inferior) is described as ‘the single greatest predictor of divorce’; defensiveness is a self-protective response to criticism that shifts blame; stonewalling is withdrawal from the interaction, usually a response to feeling emotionally flooded.
Contempt is the worst of the four horsemen. It is the number one predictor of divorce, but it can be defeated.
The Gottman Institute. The Four Horsemen: Criticism, Contempt, Defensiveness, and Stonewalling.
The popularized ‘90% (or higher) prediction accuracy’ figure attached to Gottman’s research is disputed on methodological grounds by outside commentators, and there is no publicly documented evidence of the model being validated by predicting outcomes on a fresh, unseen sample.
Reported consistently, not settled
Columbia University statistics professor Andrew Gelman documented a dispute over Gottman’s Wikipedia page, quoting journalist Laurie Abraham’s reporting: Gottman’s 1998 study knew the six-year marital outcomes of 57 newlywed couples in advance, then built a computer model to maximize the fit between videotaped communication patterns and those already-known outcomes — a retrodiction, not a prospective prediction. Abraham noted that the required next step in the scientific method — testing the equation on a new sample — was, to her knowledge, never done; each subsequent paper instead built a new after-the-fact equation. A Gottman Institute employee disputed the criticism as inaccurate but the underlying methodological point (fitting versus prospectively predicting) was not itself rebutted with new evidence in the exchange documented here.
What Gottman did wasn’t really a prediction of the future but a formula built after the couples’ outcomes were already known.
The oft-cited accuracy figures come from statistical models fit to outcomes that were already known at the time the model was built (a technique that will always look highly ‘accurate’ on the same data it was fit to). No publicly documented prospective replication on a new, unseen sample has been produced to substantiate the predictive claim as it is usually presented in popular media.
Gelman, A. (2010), quoting Laurie Abraham’s original reporting critique of Gottman’s methodology.
What to actually do
- Treat ‘contempt is the single greatest predictor of divorce’ as the Gottman Institute’s own characterization of its research program, not as an independently replicated, universally validated statistical fact — the underlying communication patterns are worth taking seriously, the precise accuracy percentages are not well substantiated.
- Watch specifically for contempt (mockery, eye-rolling, moral superiority) as distinct from ordinary criticism or complaint — Gottman’s own materials draw this distinction carefully.
What Predicts Recovery in a Troubled Marriage
The most widely circulated claim in this space — that most unhappily married people who stay together end up happy — traces to one real, specific, but non-peer-reviewed source: a 2002 Institute for American Values report by sociologist Linda Waite and colleagues, using national longitudinal survey data. The report is real and its headline finding is real, but it comes with major, frequently omitted caveats the authors themselves stated explicitly, and a peer-reviewed re-analysis of similar data three years later reached a more pessimistic conclusion about staying in a genuinely low-quality marriage.
The report’s headline finding, exactly as the authors stated it, is that about two-thirds of unhappily married adults who stayed married (rather than divorcing or separating) rated their marriages as happy five years later.
Reported consistently, not settled
This is the specific, sourced version of the widely repeated claim. It comes from a report, not a peer-reviewed journal article, published by the Institute for American Values (now the Institute for Family Studies) using data from the National Survey of Families and Households.
Two out of three unhappily married adults who avoided divorce or separation ended up happily married five years later.
The same report explicitly found that marriages marked by high conflict or domestic violence were LESS likely to become happy over time, and that divorce brought significant relief specifically in violent marriages.
Reported consistently, not settled
The authors were explicit that the ‘stay and it gets better’ pattern applied mainly to marriages made unhappy by outside stressors (illness, job loss, financial strain, a difficult child) rather than by high conflict, violence, addiction, or infidelity. This directly complicates any blanket ‘just stay together’ reading of the report.
Marriages with high conflict and domestic violence were less likely to become happy five years later.
Waite et al. (2002), Does Divorce Make People Happy?, pp. 11–14.
A peer-reviewed study using the same national survey data, published three years later and examining people in long-term (12+ year) low-quality marriages specifically, found staying unhappily married was associated with lower happiness, life satisfaction, self-esteem, and health than divorcing.
Reported consistently, not settled
This study measured more outcomes than happiness alone and focused on long-term low-quality marriages rather than the broader ‘unhappy marriage’ category in the 2002 report, and reached a notably different practical conclusion — that there was no evidence people in long-term low-quality marriages were, on average, better off than those who divorced.
There is also some evidence that staying unhappily married is more detrimental than divorcing.
The report’s own authors state repeatedly that this pattern did not hold for high-conflict or violent marriages, and that divorce offered ‘significant relief’ in violent marriages specifically. A separate peer-reviewed study of long-term low-quality marriages using similar data found the opposite pattern for that group.
Waite et al. (2002); Hawkins & Booth (2005), Social Forces, 84(1), 451–471.
What to actually do
- The ‘two-thirds became happy’ figure is real but applies best to marriages made unhappy by external stress (illness, job loss, financial strain), not to marriages marked by high conflict, violence, addiction, or repeated infidelity.
- If violence or abuse is present, the same widely cited report itself found staying is not associated with the same recovery pattern — see the abuse/coercive control topic.
Divorce Regret
This is one of the most rumor-saturated corners of the topic. Genuinely sourced academic work on divorce regret specifically (as opposed to divorce satisfaction, adjustment, or well-being generally) is thin. A commonly repeated figure — roughly a third of divorced people eventually regret the decision — traces to an academic review/thesis rather than a large peer-reviewed survey, and most of the specific percentages circulating online (54%, 73%, 80%) come from law-firm blogs, unpublished market-research surveys (e.g., Avvo.com), or social-media claims with no traceable methodology.
A BYU academic paper reviewing the divorce literature cites an estimate that roughly one-third of divorced couples eventually regret the decision within five years, alongside separate Australian research on regret.
This figure appears in an academic synthesis piece rather than as the reported result of a single large original survey; the underlying primary sources for the ‘one-third’ estimate were not independently verified in this research pass.
Barlow, B. A. (2003). Why Divorce Is Often Not the Best Option. BYU ScholarsArchive.
Most of the specific percentages circulating in articles and social media trace to law-firm marketing blogs or small, non-peer-reviewed online surveys (e.g., a 2016 Avvo.com survey of a few hundred people) rather than to a peer-reviewed academic study measuring ‘regret’ with a validated instrument. This research pass could not trace any of the frequently repeated specific percentages to a verifiable peer-reviewed source.
No verifiable peer-reviewed source found for the commonly circulated regret percentages during this research pass.
What to actually do
- Be skeptical of any specific regret percentage without a named, checkable study attached to it — this is exactly the kind of statistic the editorial brief warns is often invented.
- Frame regret honestly as understudied rather than quoting a false precision.
The Questions That Help
There is no large body of controlled research testing specific decision-clarifying questions against outcomes. What exists is a clinical framework — developed and used inside discernment counseling and the broader Minnesota Couples on the Brink Project — built around structured individual reflection (what each partner has contributed to the marriage’s decline, what has already been tried, and when the relationship has been at its best) rather than a validated ‘quiz’ or checklist with predictive research behind it.
Discernment counseling’s structure itself functions as an evidence-informed clarifying process: separate individual conversations plus structured joint sharing, focused on each partner’s own contribution to the marriage’s problems and what has already been tried.
Reported consistently, not settled
The framework does not ask ‘should you get divorced’ directly; it structures reflection on what happened to the marriage, what each partner did to contribute, what’s already been attempted, and the relationship’s best moments, before either partner has to commit to a direction.
Discernment Counseling official training site.
The Hawkins/Fackrell/Harris guidebook ‘Should I Try to Work It Out?’ is a practitioner resource built directly on the Minnesota/BYU research program, intended to help individuals and couples at the ‘crossroads of divorce’ think through the decision, but it is a clinical guidebook rather than itself a controlled study of question efficacy.
It draws on the same National Divorce Decision-Making Project survey research cited elsewhere in this file, but its recommended reflection process has not itself been validated in a randomized trial as far as this research pass could establish.
What to actually do
- Be explicit with readers that ‘good questions to ask yourself’ in this space are clinical best practice drawn from a specific, real intervention (discernment counseling), not the output of a controlled study testing questions against outcomes.
- Anchor any question list to the discernment-counseling framework (contribution to problems, what’s been tried, best moments) rather than presenting an invented list as research-backed.
What People Get Wrong
Two of the most repeated ‘facts’ in the divorce conversation have real, sourced corrections: the ‘50% of marriages end in divorce’ figure is not what current government data show, and ‘stay together for the kids’ is not supported as a blanket rule — peer-reviewed research finds it depends heavily on how much conflict is present in the marriage.
Current CDC provisional data show a much lower crude divorce rate than the ‘half of all marriages end in divorce’ figure implies, and the rate has been declining for years.
CDC’s National Center for Health Statistics reports a provisional 2023 U.S. divorce rate of 2.4 per 1,000 total population (45 reporting states plus D.C.), against a marriage rate of 6.1 per 1,000. The often-cited ‘50%’ figure is a rough, imprecise extrapolation from a different, older method of estimating divorce risk (comparing a single year’s marriage and divorce counts), not a tracked cohort outcome, and does not reflect current data.
Whether staying together ‘for the kids’ helps or harms children depends heavily on how much conflict is present in the marriage — it is not a simple, universal rule in either direction.
A national longitudinal two-part study found that the dissolution of low-conflict marriages was associated with negative effects on offspring’s later well-being (quality of intimate relationships, social support, psychological well-being), while the dissolution of high-conflict marriages was associated with beneficial effects for the children involved.
The dissolution of low-conflict marriages appears to have negative effects on offspring’s lives, whereas the dissolution of high-conflict marriages appears to have beneficial effects.
Current CDC data show a crude divorce rate of 2.4 per 1,000 population (2023, provisional) against a marriage rate of 6.1 per 1,000 — rates that have been declining for decades. The ‘50%’ figure is a widely repeated approximation, not a figure tracked directly from current government vital-statistics data.
CDC National Center for Health Statistics, FastStats – Marriage and Divorce.
Peer-reviewed longitudinal research found the opposite of a universal rule: children of parents who divorced out of a low-conflict marriage fared worse afterward than children whose high-conflict parents divorced, for whom divorce was associated with benefit. The level of conflict, not the fact of divorce itself, is what the research ties to child outcomes.
Booth, A., & Amato, P. R. (2001). Journal of Marriage and Family, 63(1), 197–212.
What to actually do
- Cite the current CDC crude rate rather than the ‘50%’ shorthand, and note explicitly that a crude rate and a cohort’s lifetime divorce probability are different measures.
- When addressing ‘staying for the kids,’ point to conflict level as the operative variable the research actually points to, rather than asserting a blanket answer in either direction.
What we could not confirm
Published because leaving it out would be the dishonest choice. Every item below is something we went looking for and could not stand behind.
- The precise percentage of divorced people who regret their decision could not be traced to any peer-reviewed, methodologically transparent study. The most commonly cited figures (roughly a third, and various figures like 54% or 80% circulating on law-firm and social-media sites) either trace to an academic synthesis piece rather than an original large study, or could not be traced to any credible source at all — treat all specific regret percentages found online as unverified unless a primary study can be produced.
- Hawkins, A. J., Harris, S. M., & Fackrell, T. A.’s guidebook ‘Should I Try to Work It Out?’ is a real, research-informed clinical resource, but its specific recommended reflection questions have not been shown in this research pass to be independently validated by a controlled study.
- The Gottman Institute’s specific accuracy percentages (frequently quoted as 90%, 91%, or 94%) for predicting divorce we could not verified as having been demonstrated on a fresh, unseen sample of couples; the only detailed methodological account found in this research pass (Gelman, 2010, citing journalist Laurie Abraham) describes the original figures as generated by a model fit after outcomes were already known.
That is a real answer and not a failure of nerve. Two things are worth doing anyway, and neither of them is a betrayal: know what your household owns and owes, and have credit in your own name. Both are things every adult should have regardless, both are useful if nothing ever changes, and both are extremely hard to arrange quickly if something does.
The most expensive mistakes in a divorce are made in the year after it — a retirement account that still names an ex-spouse, an order that divided a pension but never reached the plan, an insurance window that closed while somebody was recovering.
The checklist for after the decree → — what has a real deadline, what only looks like it does, and the three things courts have held that a decree does not do by itself.