attraction.lab

How Long Does It Take to Get Over a Breakup? Four Clocks, Four Different Answers

Three of the numbers on the first page of results trace back to one 2007 survey of 155 undergraduates, where the figure was how long ago the breakups were.

Research-backed writing on attraction, dating and relationships. Drafted by AI against a named bibliography, published only when every claim links to its source.
33 min read
Editorial illustration for: How Long Does It Take to Get Over a Breakup? Four Clocks, Four Different Answers

Before the article. If this is a bad night, you do not have to read a research summary to get through it. Find A Helpline, in the sources at the bottom, lists free and confidential services by country. In the United States you can call or text 988 at any hour. If you are being hurt, controlled or threatened by the person the relationship was with, the National Domestic Violence Hotline is 1-800-799-7233, or text START to 88788. The last section says which pattern of distress is a question for a professional rather than a question about time.

With that said. The question in the title was put to DuckDuckGo on 12 November 2026, and the eight pages of the first page of results that could be read gave between them seven different figures: eleven weeks, three months, three and a half months, six months, eighteen months, a year and a half, and half the length of the relationship. One of the eight gives the sample size of any study it describes. One links to a DOI.

Followed back, three of those numbers come from the same place. The eleven weeks and the three months are the same figure rounded differently, and in the 2007 paper they come from, that figure is how long ago the 155 undergraduates in the sample had broken up — the age of the breakups, not a measure of recovery. The three and a half months and the eighteen months come from a consumer poll run for a restaurant-review company, whose only citation is a press-release page that no longer loads. The half-the-relationship rule is attributed, by the page that uses it, to two television comedies.

Meanwhile there is a real literature, and it is unusual in that almost nobody on the first page of results has to ignore it deliberately, because it is small, old, paywalled, and mostly not phrased as a timeline. The one prospective study measured 26 people every two weeks and found that the thing they got wrong was not how fast it would fade but how bad the first day would be. The recent paper behind the "it takes eight years" headline measured something else entirely, found a half-life rather than an endpoint, and reported a decay rate whose confidence interval runs from under two years to never.

What follows is sixteen sections: what the search results contain, where each of their numbers comes from, the one page that cites well and reverses its source, then four separate clocks — distress, daily emotion, the attachment bond, and life satisfaction after divorce — what moved each of them, whether feelings can be turned down on purpose, and where the limits are.

It does not contain a number for how long this takes, because the research does not have one. Our editorial policy explains why every empirical claim here carries a linked source.

#1

What the first page of results says, counted rather than skimmed

The question in the title was put to DuckDuckGo on 12 November 2026 and cross-checked on Bing, which returned the same ten pages in almost the same order. Two of the ten would not serve themselves to an automated request — wikiHow returned a three-kilobyte stub and thehealthy.com returned a 403 — so what follows is counted over the eight that did.

The eight are Healthline, Psychology Today twice (a May 2025 piece and a January 2025 one), BetterHelp, Thriving Center of Psychology, mindbodygreen, Bustle and Grow Therapy.

  • One of the eight links to a DOI: the May 2025 Psychology Today piece, three times.
  • Two link to PubMed or PMC: Healthline, six links, and Grow Therapy, three.
  • One contains the string "et al" anywhere on the page: Healthline, seven times.
  • One gives the sample size of any study it describes: Healthline again.

The counting matters more than it sounds, because a careless version of it has already gone wrong on this site twice. Searching for "doi" in the visible text finds the letters inside "doing"; searching for "et al" finds them inside "get along". These counts are of `doi.org` inside href attributes and of "et al" with word boundaries, which is the only version worth reporting.

What the eight pages do contain is numbers: eleven weeks, three months, three and a half months, six months, eighteen months, a year and a half, and the rule that recovery takes half as long as the relationship did. Seven of the eight state at least one of these. One of the eight says where its numbers come from.

So the useful thing to do with this page is not to add a ninth number. It is to follow the ones already circulating back to the documents they came from, and then to look at what was measured prospectively, which is a different and much smaller literature.

#2

Where each number on that page comes from

Three sources account for almost all of it.

The eleven weeks, and the three months, come from one 2007 paper. It is Gary Lewandowski and Nicole Bizzoco, "Addition through subtraction: Growth following the dissolution of a low quality relationship", in the Journal of Positive Psychology. Its own abstract describes 155 undergraduates who had experienced a breakup in the previous six months, surveyed once, and the thing it was built to measure is growth: whether ending a relationship that was low in self-expansion was associated with rediscovering yourself afterwards.

Healthline, which cites it, is explicit that the eleven weeks is how long ago the breakups had happened — the average age of the sample's breakups, not a finding about recovery. It then writes that "these findings seem to imply many people recover after about 11 weeks". That inference cannot be drawn from this design at any value. A study that measures people once, at varying distances from an event, with no before-measurement and no follow-up, has no recovery duration in it to report.

The three and a half months and the eighteen months come from a press release. Healthline names it correctly as a consumer poll conducted by a market research company on behalf of Yelp, and links to it. The link goes to swnsdigital.com, a press-release wire; as of today that address does not serve a valid certificate, and the Internet Archive's retrieval service was offline when this page was written, so its snapshot from September 2021 could not be read either. The only description of this poll available to a reader following the citation is the sentence in the article citing it.

The half-the-relationship rule comes from television. mindbodygreen attributes it to Sex and the City and How I Met Your Mother, and offers no research for it. Healthline calls it a theory popularised by various media sources. Neither claims a study, which is the most accurate thing either of them does with it.

This pattern is the most reliable thing about the genre. A number appears, acquires a reason afterwards, and is then repeated by pages that would have no way of checking it even if they wanted to — which is what happened to the 3-6-9 rule and the 777 rule, to the three-day rule for texting, and to the figures attached to going no contact. Claims this site has itself had to withdraw are listed on the corrections page.

#3

The best page on the list cites the right study and reverses its result

This section exists to give credit and then to take some back, because the most interesting error on the first page of results is not in the worst page. It is in the best one.

Healthline is the only page of the eight that names a second study, gives its sample size, and describes its design. It writes: "Another 2007 study aimed to compare the level of distress people thought they might experience after a breakup with the actual distress they experienced. Of the 69 participants, 26 experienced a breakup within the first 6 months of the study. These participants reported their distress by filling out a questionnaire every 2 weeks."

That is accurate, and the study is the right one — it is the only prospective study of this question in the literature, and the next sections are about it. The link Healthline gives for it goes to the authors' own copy of the paper, which is readable.

The sentence that follows is where it goes wrong: "Their distress declined steadily over several weeks, just as they had predicted, and by the 10-week mark, they felt better."

Half of that is right. The participants' distress did decline at about the rate they had predicted. But the paper's finding, stated in its title and abstract, is that the forecast was wrong — and precisely about the part Healthline's sentence drops. Predicted distress was significantly higher than actual distress from the first measurement onward, and it was still significantly higher at week ten, the last measurement taken.

The difference is not pedantic. If people get the fade right and the peak wrong, then "you have a long haul ahead" is the wrong reassurance and "it will not be as bad as you are picturing" is the right one. A gloss that says people predicted this accurately removes the only practically useful thing the study found.

#4

"Getting over it" is four different things, on four different clocks

Before any number is worth quoting, it is worth asking what it is a number of, because the four research literatures that touch this question are measuring four different objects and reporting them on four different timescales.

Self-reported distress. How upset you say you are. Measured prospectively in exactly one study, over ten weeks.

Day-to-day emotion. Love, sadness, anger and relief sampled repeatedly over the weeks after the ending, which gives a shape rather than an endpoint.

The attachment bond. Whether the other person is still someone you would turn to. The name this literature gives to the change is attachment reorganization, from a 2008 model by David Sbarra and Cindy Hazan; the description used here is the one in Sbarra and Jessica Borelli's later review, which is the version we read. The endpoint is defined carefully: a shift from a state in which your sense of security is regulated partly by one specific person to one in which it is not, at which point reminders of the loss stop being events that take substantial effort to get through. That is not the same as feeling nothing. It is the point at which feeling something stops costing you.

Life satisfaction and mental health in panel data. Measured in national surveys over years, almost always for divorce rather than for breakups.

These do not have to move together, and the evidence says they do not. Somebody can be sleeping fine, functioning at work and no longer crying, and still answer yes to "I would like to be able to talk to them". The genre treats that combination as a failure of recovery. On the measurement side it is just two clocks reading different times, which is what you would expect them to do.

The four sections that follow take each clock in turn, with what was measured, on whom, and for how long. Three of the four are small. One is large and is about divorce. None of them produces the number the search box is asking for.

#5

Clock one: the only study that asked before the breakup and after

Paul Eastwick, Eli Finkel, Tamar Krishnamurti and George Loewenstein published it in the Journal of Experimental Social Psychology in 2008, and the design is the reason it is the only one.

They recruited 69 Northwestern freshmen, all aged 17 to 19 and all already in a dating relationship of at least two months, and sent them an online questionnaire every other week for 38 weeks. While still together, participants predicted how they would feel two, four, eight and twelve weeks after a hypothetical breakup. Thirty-eight percent broke up during the following six months, and when they did, they reported their actual distress at the breakup session and again two, six and ten weeks later.

That is what makes it rare: the forecast on record was made two weeks before the breakup was reported, by the same person who then lived through it. Every other study of this question either asks people to predict a hypothetical, or asks people who have already broken up to remember.

The analytic sample is 26 people. That is the whole study, and it is the main thing to hold against everything below.

The results, in the paper's own coefficients, on a one-to-seven scale:

  • Predicted distress exceeded actual distress at the moment of the breakup, by 0.79 of a scale point (p < .001).
  • Actual distress declined over the ten weeks, by about 0.07 of a point per week (p = .021) — roughly seven tenths of a point in total.
  • Predicted distress declined at a rate that was not statistically distinguishable from the actual rate (p = .591).
  • At week ten, prediction still exceeded reality by 0.64 of a point (p = .009).

So the honest summary of the only prospective measurement is: distress was lower than expected from the first day, fell slowly and steadily, and was still falling when the study stopped looking.

#6

What an initial-intensity error means for the person waiting it out

The authors frame their result as an argument between two explanations of why people mispredict their own feelings. Either you are bad at guessing how intense the first moment will be, or you are bad at guessing how fast it fades. Their data say the first and not the second, cleanly.

That distinction is the practical content of the paper, and it points the opposite way from the reassurance this genre usually offers.

The standard consolation is about the fade: it gets better, give it time. The measured error was not there. People in this study were approximately right about how fast they would improve. What they were wrong about was how bad the worst of it would be — they expected the floor to be lower than it turned out to be.

The paper also found who made the error. Participants were asked, at the session before the breakup, how much they were in love. Those a standard deviation above the mean overestimated their distress by a full scale point. Those a standard deviation below it were accurate, with a coefficient of −0.15 that was nowhere near significance.

Read carefully, that is not a story about resilience. It is a story about what happens when you forecast from inside a strong feeling: the feeling supplies the estimate. The people who were not especially in love had nothing distorting the arithmetic, and they got it right.

The same error shows up on the other side of the decision. In the research on staying, people asked to imagine leaving also overestimate how bad it will feel — and the overestimate is largest among the people least able to leave, which is covered in why people stay in relationships that are not making them happy. It is also worth separating the intensity of a feeling from what it indicates, because intensity and attachment anxiety are easy to confuse with love and the confusion changes what you predict about yourself.

Two limits before this travels anywhere. The sample is 26 American first-year undergraduates who had been dating for at least two months, which is close to the smallest and youngest population that could have produced this question. And the follow-up ended at ten weeks, so the paper has nothing at all to say about month six or year two. It is a study of the first stretch, not of the distance — which, given that the first stretch is the part people mispredict, is a reasonable place to have aimed.

#7

Clock two: the daily diaries, where love falls in a straight line and sadness does not

The second literature samples emotion repeatedly instead of asking for a summary, and it belongs almost entirely to David Sbarra and colleagues at Arizona.

In the 2005 study with Robert Emery, published in Personal Relationships, 58 young adults were recruited while they were still in a serious dating relationship, and when it ended they were signalled at random by beeper for 28 days to fill in an emotions diary. A comparison group in intact relationships supplied the baseline.

Four findings from the paper's abstract, which is as far as this page can go because the full text is behind a paywall:

  • Compared with people whose relationships were intact, the dissolution group showed more emotional volatility, and most of all immediately after the breakup.
  • Love declined linearly. Sadness, anger and relief followed curvilinear patterns.
  • Contact with the former partner slowed the decline for both love and sadness.
  • Attachment style and the rated impact of the breakup predicted both the starting points and the rates of change.

That last one is the only place in this literature where a stable individual difference does visible work on the timeline, and it is worth knowing that the difference is itself not fixed — whether attachment styles change, and by how much is a question with a measured answer rather than a reassuring one.

The first two are the ones that break the usual framing. Volatility is highest when distress is highest, which means that in the period when people most want a trend line they are living in the part of the data with the most scatter. And the four emotions do not travel together: love running down a straight slope while sadness curves is a measured description of the state the genre calls mixed feelings.

The third finding has a page of its own on this site, because it is the only piece of evidence in this literature that looks like an instruction — and it is weaker and narrower than the rule built on top of it, as the research behind the no-contact rule sets out. Slowing the decline of love and sadness is not the same as making anybody better off, and this study did not test that.

#8

The one study that modelled recovery as a time-to-event, and why this page still gives no median

If a median recovery time exists anywhere in this literature, it is in Sbarra's 2006 paper in Personality and Social Psychology Bulletin, and it is the only serious candidate.

It is built from the same kind of data: daily emotion reports for four weeks from 58 young adults whose serious dating relationships had recently ended. What makes it different is the analysis. Using baseline data from people in intact relationships to define what a recovered level looks like, Sbarra defined sadness recovery and anger recovery as events in time and modelled them with Cox survival analysis — the statistical machinery built precisely to answer "how long until".

From the paper's abstract, the predictors behaved as follows. Acceptance that the relationship was over mediated the link between attachment security and sadness recovery. Higher levels of love, anger and attachment preoccupation were associated with a decreased probability of recovering from sadness during the study window. Attachment security was associated with an increased probability of anger recovery, while ongoing sadness decreased it.

The paper reports that sadness and anger recover differently and on different schedules, which is a real finding and the reason a single number for "getting over it" was always going to be a compromise between two unlike things.

And then this page stops, because the article is paywalled and its abstract does not state a median. Three routes were tried: the publisher, the open-access indexes, and the author's institutional repository. The only full-text copies findable are on piracy mirrors, which are not used here.

So the position is worth stating plainly. The single study in this field designed to produce a time-to-recovery figure is one this page cannot read, and therefore no median appears above. If a number does exist in that paper, it describes 58 young adults over four weeks, which is a narrower thing than any of the figures currently circulating.

Further illustration for: How Long Does It Take to Get Over a Breakup? Four Clocks, Four Different Answers
#9

Clock three: the attachment bond, and the paper behind the "eight years" headline

The second-ranked result for this question is a Psychology Today piece from May 2025 that opens by promising a definitive answer, and it is built on a real and recent paper.

The paper is Jia Chong and R. Chris Fraley, "The Long-Term Stability of Affective Bonds After Romantic Separation: Do Attachments Simply Fade Away?", published online in Social Psychological and Personality Science in March 2025 and now in volume 17, issue 1. It is paywalled, but it is the published version of Chong's 2022 doctoral dissertation at the University of Illinois, which is openly available in that university's repository and contains the same study, the same sample and the same tables. Everything below is read from there.

What it measured is not distress. Participants rated how much they would want to use a former partner for the three classic attachment functions — being near them, turning to them when upset, relying on them as a secure base — on a 0 to 4 frequency scale, with items like "Whether or not I actually interact with them, I would like to be able to see or talk to my former partner".

Who was in it. 320 people, 57% women, mean age 32, recruited from a university credit pool and Amazon Mechanical Turk between March and July 2022. To qualify, a person had to have had a past relationship lasting more than two years and to no longer be in it. On average that relationship had lasted 4.7 years and had ended 5.1 years before the survey.

How it was measured. One survey. Everybody answered once. Time since the breakup is a difference between people, not a trajectory within anybody, and nobody in this study was followed over time.

The finding is a curve: ratings of former partners start high and decay toward an asymptote, and the asymptote is statistically indistinguishable from how people rate a stranger or an acquaintance they are not close to.

#10

4.18 years is a half-life, and a half-life does not double

The number that travelled is in the published abstract: attachment bonds are relinquished with "4.18 years as a mid-point". Psychology Today turned that into "on average, it can take up to eight whole years to well and truly be over an ex-lover".

That does not follow, for three separate reasons.

First, doubling a half-life does not finish the job. The model is exponential decay toward an asymptote. One half-life gets you halfway there; two half-lives get you to three quarters, not to the end. The curve approaches its floor and never reaches it. Multiplying 4.18 by two produces a number that corresponds to nothing in the model.

Second, the floor is not zero, and the paper says so in its title. The dissertation's own sentence is that the bond "fades to a point of insignificance", because the asymptote sits at the level people give a stranger. In the pilot, strangers and non-close acquaintances scored 1.03 on the same 0-to-4 scale — people will mildly endorse these items about almost anybody. Former partners settle at that level. The paper's summary of its own result is that "people's former partners simply become someone they used to know", which is an argument that the bond ends, not that it lasts eight years.

Third, it is not a measure of being over someone in the sense the headline implies. Wanting to be able to talk to an ex is not heartbreak. The construct is whether a specific person still occupies an attachment role, and plenty of people reach the stranger-level asymptote while remembering the relationship warmly.

One arithmetic note of my own, offered as arithmetic rather than as a correction. In the open dissertation the decay constant is 0.19 per year, and ln(2) ÷ 0.19 gives 3.6 years — which matches the 44.10 months the dissertation reports for one version of the model, but not the 50.11 months it reports for the other, even though both versions share that same constant and a half-life does not depend on where you put the zero. The published paper carries the larger figure. I cannot read it to see whether it was refitted, and for any practical purpose 3.6 and 4.2 years are the same answer.

#11

How uncertain that number is, from the same table

A more serious issue sits two columns to the right of the figure everyone quotes, and it is printed in the dissertation's own tables.

The decay constant is reported as 0.19, with a 95% confidence interval from −0.06 to 0.44, and p = .14. In the stranger-centred models it fails to reach significance in all seven specifications — the p-values run from .08 to .27.

Translate that interval into the units the headline uses, which is a monotone conversion and therefore legitimate. At the top of the interval the half-life is about 19 months. At the bottom it is negative, meaning the data are consistent with no decay at all over the range observed. The point estimate of roughly four years sits inside a range that stretches from under two years to never.

What the study does establish firmly is the asymptote, not the speed. The asymptote is where the paper's argument lives and where its statistics are strong: ratings of former partners settle at a level that does not differ from ratings of strangers, which is a genuine answer to the theoretical question the paper set out to settle — do old bonds persist in some reduced form forever, or are they given up?

The speed at which people get there is the part the design is least able to see, and the authors say so. Their own limitations section names it: the sample was self-selected into a study about breakups, and "a longitudinal study would also allow for a more precise time estimate".

There is a smaller tell in the same tables, worth noting because it is visible without any statistics training. In the models that test whether attachment style changes the rate, four standard errors come back as 110, 196, 325 and 520 — on coefficients that are themselves between −7 and 18. Numbers like that mean the model could not estimate those terms at all. They are not findings in either direction, and the author reports them as non-significant rather than dressing them up.

#12

What that study's sample does and does not cover

Three features of who was in it decide where the four-year figure can and cannot be carried.

Everybody's relationship had lasted more than two years. That was an eligibility rule. So the figure says nothing about a four-month relationship, a situationship, or anything that ended before a bond of this kind had formed — and the question of whether one formed at all is the harder one when what ended was never defined, which the research on undefined relationships goes into.

The typical participant was five years past the breakup. This is a sample of people looking back, recruited on the internet to talk about an old relationship, not a sample of people in the acute weeks.

Nobody was followed. The curve is drawn across people, so it assumes that somebody six years out is a fair picture of where somebody two years out will be. That assumption is doing a lot of work and it is untestable in this design.

The moderator results are worth having anyway, because they are the parts that bear on the questions people actually ask.

  • Attachment anxiety went with more wanting to be near an ex and turn to them when upset; avoidance went with less. Neither changed the shape over time. Those two dimensions — anxiety and avoidance — are what our attachment style test scores, and it is an educational tool rather than a diagnosis or a validated clinical instrument.
  • 36% had initiated the breakup, and initiating made a difference to one function only — wanting to be near them — and not to the rest.
  • 58% had been in a new relationship since, and that was associated with essentially nothing. No main effect, no interaction, one small exception on one composite. Whatever a new relationship does, in this dataset it did not shorten the bond to the old one.

One reading of the measure is worth making explicit, because it decides what the finding means. The items are about wanting to be able to see, talk to and count on the other person. That is close to what people are describing when they ask whether staying friends with an ex is possible or advisable, and this study says the wish itself is ordinary and declines slowly rather than being a sign that something has gone wrong.

#13

Clock four: divorce, where the average hides at least three different trajectories

The divorce literature is the only one here with real statistical power, and it reaches a conclusion that works against the premise of the question.

Richard Lucas analysed an 18-year German panel of more than 30,000 people in Psychological Science in 2005. Life satisfaction drops as divorce approaches and then rebounds — but, in his words, "the return to baseline is not complete". The same analysis found that people who would later divorce were already less happy than people who would not, before either group married. Some of what looks like the cost of a divorce was there in advance of the marriage. That paper does more work in why people stay in relationships that are not making them happy, where its two halves matter in a different way.

Then the averages were taken apart. Anthony Mancini, George Bonanno and Andrew Clark applied latent growth mixture modelling to panel data on bereavement, divorce and marriage, and found that divorce fits a three-class solution rather than one curve. Their headline is the one that should end any conversation about an average timeline: the modal response across these events was a relatively flat trajectory. The most common thing that happened to people's subjective well-being was not very much.

The most detailed picture comes from Switzerland. Bina Knöpfli, Davide Morselli and Pasqualina Perrig-Chiello followed 308 people who divorced after marriages averaging 23.6 years, mean age 55.6, and tracked them across two years with latent transition analysis. Five groups: resilients 29%, average copers 49%, vulnerables 6%, malcontents 12%, resigned 4%. Three quarters were doing well and a fifth were not.

And the finding that matters most for a page about time: across those two years, only a small number moved from the poorly adapted groups to the well adapted ones. The copers and the resilients stayed where they were. Within this window, the trajectory somebody was on predicted where they would be better than the passage of time did.

#14

What moved the clock in the studies that measured it

Four things in this literature have been tied to recovery with something better than assertion, and one of them is strange enough to be worth the space.

Getting your sense of yourself back. Erica Slotter, Wendi Gardner and Eli Finkel showed across three studies that a breakup changes both the content of the self-concept and its clarity, and that reduced clarity uniquely predicted post-breakup distress — over and above how upset people were about the relationship itself. The thing that hurts is partly not knowing who you are now. The same construct turns up in an unrelated corner of this site, in what happens to people who shape themselves around what others want, which is a reason to think it is doing real work rather than decorating one finding.

And then: being asked about yourself, repeatedly. Grace Larson and David Sbarra took 210 recently separated young adults and assigned them either to four visits of intensive assessment over nine weeks or to an intake and an exit only. The intensively measured group reported larger decreases in self-concept disturbance, and that improvement explained decreases in intrusive thoughts about the breakup, in loneliness, and in how often they said "we" when describing it. No other main effect appeared. Answering questions about yourself every few weeks, with no therapeutic intent and no expectation set, moved the thing that predicts distress. It is also a warning about every diary study above: measuring this appears to change it.

Having an account of it that hangs together. Kyle Bourassa and colleagues found that how coherent people's narrative of their separation was — whether they felt they understood the story of it — predicted lower psychological distress 7.5 months later. Sbarra and Jessica Borelli's review treats narrative coherence and self-concept clarity as the two trackable markers of reorganization.

What did not move it. Entering a new relationship, in the 320-person study above, was associated with essentially no change in the bond to the previous partner — a result that sits awkwardly against the one empirical study of rebounds, from the same senior author eleven years earlier, as whether rebound relationships work sets out. And contact with an ex, in the diary data, slowed the decline of love and sadness rather than speeding it.

#15

Can you decide to feel less? What was tested, and at what size

There is one laboratory literature on deliberately turning the feeling down, and it is small enough that its size has to be stated before its result.

Sandra Langeslag and Jan van Strien published it in PLOS ONE in 2016, open access. Two parts. First, a questionnaire study asking people whether love feelings can be controlled at all, and what they do when they try. The answer to the first was that people perceive love feelings as somewhat uncontrollable. The answer to the second was a list that will be familiar to anybody who has done it: reappraisal, distraction, avoidance, and taking up new activities.

The second part tested whether the belief is right. Participants were cued to think about their partner or former partner in ways designed to raise or lower the feeling, and the effects were measured both by self-report and by an EEG index of how much attention the person was giving to the cue.

Instructed down-regulation using reappraisal decreased self-reported infatuation and attachment. The electrophysiological measures moved in ways that partly supported this and partly did not — up-regulation produced one effect at the short latency and an inconsistent one at the long latency, which the authors report as a complication rather than smoothing over.

And the sample: 20 people in a relationship and 20 who had recently broken up, aged 19 to 26, recruited at one Dutch university, with the brain measures resting on 18 and 18 after exclusions. That is a pilot-scale study. A follow-up exists in the Journal of Experimental Psychology: General, and it is paywalled, so its numbers are not quoted here.

What survives is modest and still worth having: the belief that feelings are not controllable is itself testable, it has been tested once at a small size, and it did not hold up. Nothing here supports a technique, a timeline, or a promise that deciding to feel less works reliably.

It is worth saying what this is not, because the nearby advice market sells the opposite. There is no evidence here for making yourself stop caring on a schedule, and none for doing it to anybody else. The pattern of attaching hard and fast to people regardless of who they are has its own measured literature, in the early signs of emophilia, and that is a different question from whether a particular feeling can be turned down once.

#16

What none of this can tell you, and when slowness is not the problem

Put the four clocks side by side and the honest answer to the question in the title is that it depends on which one you are reading, and that three of the four were measured on very few people.

Twenty-six American first-year undergraduates, prospectively, for ten weeks. Fifty-eight young adults with beepers, for four weeks, twice. Three hundred and twenty internet and undergraduate participants, once each, all of them remembering a relationship that had lasted more than two years. Against that, two national panels and a Swiss cohort, all of them about divorce after a marriage, mostly in middle age, and mostly in German-speaking Europe.

Almost none of this is about same-sex couples. None of it is about a relationship that ended after six weeks. The one study built to produce a time-to-event figure is one this page could not read.

What can be said with the evidence in hand is smaller than a number and more useful than one. The worst of it is probably not as bad as you are currently forecasting, and that is the measured error. The improvement is real but slower than any of the circulating figures imply, and in the only prospective data it was still going at the ten-week mark. The emotions do not move together, so feeling mixed is not a sign of being stuck. And the point at which an old relationship stops occupying the role it had is better described as the point where reminders stop costing you effort than as the point where you feel nothing.

One pattern is different in kind from slowness, and it is the one worth taking to a professional rather than to a timeline: not fluctuation, but flatness — distress that is not moving at all over months, or getting steadily worse, or coming with thoughts of harming yourself. That is not a clock running slow. The same signal, read before the ending rather than after it, is what the longitudinal data on thinking about breaking up is about, and what has been measured about having moved on covers the markers rather than the duration. The rest of this section is under Mindset.

Pulling it together

The short version is that the question has four answers because it is four questions, and that the numbers in circulation are a sample descriptor, a press release and a sitcom.

What was actually measured: in the only study that recorded a forecast before the breakup and the reality after it, 26 people overestimated their distress by about four fifths of a scale point on day one and were still overestimating it ten weeks later, while being roughly right about the rate of improvement. In diary data from 58 young adults, love declined in a straight line while sadness, anger and relief curved, and contact with the ex slowed the first two. In 320 people looking back at relationships that had lasted more than two years, the desire to turn to a former partner decays to the level people report for an acquaintance — with a half-life estimated at about four years and a confidence interval running from under two years to no decay at all. And in panel data on divorce, the single most common trajectory for life satisfaction was a flat one, with a fifth of people in a Swiss cohort doing badly two years on and few of them moving groups.

Three things seem durable across all of it. The peak is the part people misjudge, not the fade. Getting a clear sense of yourself back tracks recovery better than elapsed time does, and it moved when people were simply asked about themselves every few weeks. And the end state is not the absence of feeling; it is the point at which a reminder stops taking effort.

If what is happening to you is not slow but flat — no movement at all over months, or a steady worsening — that is the pattern worth taking to someone, and it is not a question about timing.

Frequently asked questions

How long does it take to get over a breakup?

There is no measured figure to give, and the figures circulating are not what they appear to be. The only prospective study — Eastwick, Finkel, Krishnamurti and Loewenstein, who tracked 69 university freshmen every two weeks for 38 weeks and captured 26 breakups — found distress declining by roughly 0.07 of a point per week on a seven-point scale, still falling when the study stopped at ten weeks. Diary studies of 58 young adults found love declining linearly and sadness curvilinearly over four weeks. A 2025 study of 320 people found that the wish to turn to a former partner decays toward the level people report for an acquaintance, with a half-life estimated at about four years but a confidence interval stretching from under two years to no decay at all. These measure different things on different populations, which is why they do not combine into one answer.

Where does the "11 weeks" figure come from?

From a 2007 paper by Gary Lewandowski and Nicole Bizzoco in the Journal of Positive Psychology, where it is not a finding. That study surveyed 155 undergraduates once, all of whom had broken up within the previous six months, and its subject was growth after ending a low-quality relationship. The eleven weeks is the average time that had passed since those breakups — a description of the sample, not a measurement of how long recovery took. Healthline, which cites it, states this correctly and then writes that the findings "seem to imply" recovery at about eleven weeks. They cannot imply it: a study that measures people once, at varying distances from an event, with no baseline and no follow-up, contains no recovery duration at all. The "three months" figure that appears elsewhere on the same search results page is the same eleven weeks, rounded.

Is it true that it takes half the length of the relationship?

There is no research behind it. The page on the first results page that states the rule most plainly, mindbodygreen, attributes it to Sex and the City and How I Met Your Mother; Healthline calls it a theory popularised by media sources. Neither claims a study, and no study of breakup recovery known to us models recovery time as a proportion of relationship length. The nearest relevant evidence runs the other way: in a 2025 study of 320 people, length of the former relationship was not what distinguished the trajectories, and in a Swiss study of 308 people divorcing after long marriages, length of marriage played only a marginal role in which adaptation group someone belonged to. Personality, finances and whether there was a new relationship mattered more.

Does it really take eight years to get over an ex?

No, and the study behind that headline does not say so. Chong and Fraley (2025) modelled the decay of attachment bonds to former partners in 320 people and reported 4.18 years as a mid-point — a half-life. A half-life does not double into completion: two half-lives leave you three quarters of the way, and the curve approaches its floor without reaching it. The floor also is not zero. It sits at the level people report for a stranger or an acquaintance, which is why the authors summarise their own result as former partners becoming "someone they used to know". The study measured whether an ex still occupies an attachment role — whether you would still like to be able to talk to them — not heartbreak. And the decay rate it rests on had a confidence interval from −0.06 to 0.44, so the data are consistent with a half-life anywhere from about nineteen months to no decay at all.

How long does it take to get over a divorce?

The panel evidence says the average is the wrong statistic. Richard Lucas, using an eighteen-year German panel of more than 30,000 people, found life satisfaction dropping as divorce approached and then rebounding without fully returning to baseline — and also found that people who would later divorce were less happy before they ever married. When Mancini, Bonanno and Clark broke similar data into trajectory classes, the most common response to divorce was a relatively flat trajectory: not much change at all. In the most detailed follow-up, 308 Swiss adults who divorced after marriages averaging 23.6 years split into five groups two years on — 29% resilient, 49% coping averagely, and 22% across three struggling groups — and very few moved from the struggling groups to the thriving ones within those two years. The eighteen-month figure in circulation comes from a consumer poll, not from any of this.

Does no contact make you get over someone faster?

The one measurement that bears on it is narrower than the rule built on it. In Sbarra and Emery's diary study of 58 young adults beeped at random for 28 days, contact with the former partner slowed the decline of both love and sadness. That is a real association, and it is the empirical seed of the whole no-contact genre. What it does not establish is that avoiding contact makes anyone better off: slowing the decline of love is not harm, the study was not an experiment, and nobody was assigned to a contact condition. There is also no evidence anywhere in this literature for a specific duration — the thirty-day and sixty-day rules have no measured basis. We go through what the research does and does not support about that rule separately.

Does getting into a new relationship help you move on?

In the only study that tested it against the bond itself, no. Chong and Fraley asked their 320 participants whether they had been in a committed relationship with somebody new since the breakup; 58% had. There was neither a main effect nor an interaction with time on how much they still wanted to use their former partner for attachment functions, with one small exception on one composite. That is a cross-sectional finding in a self-selected sample and it cannot rule out benefits elsewhere — on loneliness, on mood, on day-to-day life. What it does say is that a new relationship did not visibly shorten the old bond, which is the thing a rebound is popularly supposed to do. The rest of that evidence, including the two experiments that never put anybody in a relationship, is in what the research on rebound relationships actually measured.

Is it normal to still miss my ex after a year?

On the measurement evidence, yes, and the framing of it as a failure comes from nowhere empirical. In the 320-person study, the typical participant was five years past a relationship that had lasted nearly five years, and the modelled curve still sat above the stranger baseline at a year. The endpoint that attachment researchers describe is not the absence of feeling. In Sbarra and Hazan's model, as summarised in Sbarra and Borelli's review, successful reorganization is the point at which your sense of security no longer depends on one specific person and reminders of the loss stop being events that take substantial effort to get through — a change in cost, not in memory. Missing somebody occasionally after a year is consistent with having reorganized. Distress that is not moving at all over months is a different pattern, and that one is worth taking to a professional.

When should I talk to a professional about a breakup?

The signal is not slowness. Recovery in this literature is volatile, especially early, so a bad fortnight after a good one is an expected feature rather than evidence of being stuck. The patterns that warrant attention are flatness — no movement at all across months — distress that is steadily worsening rather than fluctuating, an inability to function at work or with other people that is not easing, or any thought of harming yourself. In the Swiss divorce cohort, about a fifth of people were in groups that were adapting poorly two years on, and very few of them moved out of those groups on their own within the study window, which is a reason to treat persistent difficulty as something to get help with rather than something to wait out. In the United States you can call or text 988 at any hour; Find A Helpline, in the sources below, lists services by country.

Sources

What this article draws on. For the broader bibliography see our sources page.

Keep reading