How Long Does Limerence Last? Two Years, Halved and Doubled

Limerence typically lasts about two years. The largest study of it to date, covering 1,647 people and published by Chloe Evans and colleagues in Acta Psychologica in July 2026, describes episodes involving roughly two years of obsessive fixation on a single person. Dorothy Tennov, the psychologist who named the state in her 1979 book Love and Limerence, put the usual span at 18 months to three years. In an earlier University of Sheffield doctoral survey of 235 people with limerent experience, the median shortest episode reported was six months and the median longest was four years. Episodes running a decade or more appear in that same data.

Where the two-year number actually comes from

Three separate bodies of evidence land near two years, built in completely different ways.

Tennov's figure came from about a decade of collecting: thousands of questionnaires, several hundred interview case studies, plus autobiographies and published journals. It is a descriptive summary of a qualitative corpus, repeated for forty-seven years without ever being replicated as a measurement. Worth knowing before you treat it as a statistic.

The Sheffield doctoral study, run by Chloe Evans, put distributions on it. Two hundred and thirty-five people across 34 countries were asked two blunt questions: what is the shortest amount of time a limerent episode has lasted, and what is the longest. Shortest episodes ranged from five minutes to 30 years, mean 15.1 months, median six months. Longest episodes ranged from 95 minutes to 40 years, mean 61.0 months, median four years, standard deviation 65.57 months.

Read that last cluster again. The standard deviation is larger than the mean. When spread exceeds centre, the average has stopped describing anybody.

I learned to respect that signal the expensive way. In 2011, keeping inventory for a regional food bank, I published a per-case yield for donated canned vegetables: 11.4 servings, a mean across everything that came through the dock that quarter. Three pallets of institutional restaurant-size donations sat inside that average. A partner pantry built a six-week meal plan on 11.4 and ran short in week three. The median case yielded seven. I spent the following year walking back a number I had already printed, and I have not published a mean without its median since.

Limerence duration has that shape. A handful of people reporting 30 and 40 years drag the mean away from where most respondents sit. The median is the number to hold.

The 2026 Acta Psychologica paper then arrived with 1,647 people and reported roughly two years of fixation per episode, an onset around age 18, and about five episodes recurring across adulthood. Two years is where a 1979 qualitative summary, a 235-person distribution and a 1,647-person survey converge.

An easy way to hold the figure: two, halved twice, doubled once

Here is the whole method:

1. Start at two years. That is the central estimate from the largest study, and it sits inside Tennov's original 18-month-to-three-year window.
2. Halve it twice for the short end. Two years becomes one year becomes six months. Six months is the median shortest episode in the Sheffield data.
3. Double it once for the long end. Two years becomes four years. Four years is the median longest episode in the same data.

Six months, two years, four years. Three numbers, one arithmetic move each, and every one a published median rather than something I rounded into shape.

Be clear about what the ladder is not. The six-month and four-year figures are medians of two different survey questions, not quartiles of one distribution. They mark where the typical person's briefest and longest episodes fall. They do not mean 50% of episodes fit between them.

Until about 2019 I refused to lead with a single figure for anything skewed this hard. I published the range and let readers place themselves in it, which felt honest. I stopped because of what readers did with it. Handed "five minutes to 40 years," someone already frightened by their own preoccupation reliably anchors on 40. A range with no centre quietly hands the reader the worst number in it. So I lead with the centre now and derive the range from it, the reverse of what I used to do.

How much of the day does limerence take up?

The 2026 paper's second study measured this in real time rather than from memory, a first. Fifty-one people carrying impairing limerent episodes were sampled repeatedly across seven days. Thoughts about the limerent object occupied 50% of waking thought, and those thoughts were intrusive, immersive and tightly linked to negative mood. Intrusive-distracting fantasy was the dominant pattern, surging during sadness and loneliness.

The doctoral run that preceded it gives the mechanics. Sixty-two participants were texted eight times a day for seven consecutive days, at random moments between 09:00 and 22:00, with a 30-minute response window. That produced 2,073 usable responses at a 60% response rate. Mind-wandering about the limerent object came in at 45% of sampled moments, standard deviation 26 points, individual range running the full width from 0% to 100%.

Watch the unit. "Half of waking thought" is a proportion of sampled moments, not half the clock. Nobody measured hours. The limerence-adapted Yale-Brown scale used in the same research does ask about hours, in bands: none, under one a day, one to three, more than three and up to eight, or more than eight and near-constant. Two instruments, two units, one topic. Mix them and you will tell someone they spend twelve hours a day thinking about a coworker when the data said no such thing.

This is where my usefulness stops and I would rather say so. I have never been limerent. I cannot tell you what month 14 feels like from inside, and anyone deriving that from a spreadsheet is selling something. What I can vouch for is what happens to a duration figure when the denominator changes, when the sample selects itself, and when mean and median disagree. That is my trade, and it applies here as cleanly as it applies to case weights.

Limerence vs. infatuation: what each clock actually measures

Ordinary early attraction has its own timer, close enough to limerence's that people conflate the two.

Donatella Marazziti's team in Pisa published the comparison that made the point in 1999, in Psychological Medicine. They recruited 20 people who had fallen in love within the previous six months, 20 unmedicated patients with obsessive-compulsive disorder, and 20 controls, then measured platelet serotonin transporter density. The newly in love were indistinguishable from the OCD patients, and both sat significantly below controls. Early romantic attraction, biochemically, looked like an obsessive-compulsive state.

The follow-up matters more here. In 2004, in Psychoneuroendocrinology, Marazziti and Canale measured pituitary, adrenal and gonadal hormones in 24 people who had recently fallen in love against 24 comparison subjects. Cortisol ran higher in the in-love group; men showed lower FSH and testosterone, women higher testosterone. Every one of those differences was gone when the same subjects were retested 12 to 24 months later.

So infatuation's measurable signature clears inside one to two years. Limerence's central duration estimate is about two years. The clock does not separate them.

| | Early infatuation | Limerence | Impairing limerence | |---|---|---|---| | Typical span | Hormonal differences cleared by 12–24 months (Marazziti & Canale, 2004) | ~2 years per episode (Evans et al., 2026) | ~2 years, recurring ~5 times across adulthood | | Sample behind the figure | 24 in love vs 24 comparison | 1,647 surveyed | 51 sampled in real time over 7 days | | Reciprocation | Not part of the definition | Absence of it is the definition | Absence of it is the definition | | Thought load measured | Not measured by moment sampling | 45–50% of sampled waking thought | Same, plus intrusiveness and negative affect | | Comorbidity reported | None reported | Raised adverse childhood experiences, insecure attachment | 42% simultaneously reaching caseness on four measures |

What separates them is the middle rows. Infatuation does not require the other person to stay out of reach. Limerence is defined by unreciprocated longing, and the paper's impairing form adds a comorbidity load ordinary attraction does not carry: 42% of limerent respondents simultaneously reached caseness for anxiety, depression, dissociation and maladaptive daydreaming.

How long does limerence last with no contact?

Nobody has measured it. No study has tracked a no-contact group against a contact group with a clock running, and any article giving you a confident figure has invented one.

What exists is adjacent. The Sheffield survey asked 205 respondents who their limerent object was. A friend led at 26.8%, then a co-worker at 20.0%, an acquaintance at 13.2%, a stranger at 11.2%, an ex-partner at 9.8%. Nearly half, 46.8%, are friends or co-workers. For those people "no contact" is not a decision. It is a job change or the loss of a friend group, and the researchers flagged actual contact volume as an open question their own data could not answer.

Tennov supplies the theoretical prediction. She argued that uncertainty sustains limerence, and that intermittent reciprocation, enough to keep hope alive but never enough to settle it, stretches an episode out. If that is right, removing the intermittent signal should shorten the episode. Reasonable prediction, untested against a calendar.

The one published treatment record is not about avoidance at all. Brandy Wyant's 2021 case study in the Journal of Patient Experience documents exposure and response prevention, the technique used for OCD, applied to a 28-year-old with limerent attachments dating back to age four. Her current limerent object was a coworker, and she was ruminating several hours a day. Across matched two-week self-monitoring windows, compulsive rituals fell from 225 occurrences to 10, and time on them from about eight hours to ten minutes. On a 60-point screening instrument she developed, her score moved 43 at day zero, 38 at day 12, 30 at day 30, 23 at day 49. Sessions ran 45 minutes every two weeks, and improvement held at nine-month follow-up.

Read that with the caveat it deserves: n equals one, the diagnosis was self-made, and the patient was the author. One case is not an effect size. It is also the only outcome data in print.

What are the stages of limerence?

Tennov described three: infatuation, then crystallization, where the person is idealised and thoughts turn intrusive, then deterioration, where the intensity quiets and the idealisation dissolves. She also described three exits, which people confuse with the stages. Consummation is reciprocation, which can convert the state into something better called love. Starvation is hope running out. Transference is the fixation relocating to someone new.

Albert Wakin and Duyen Vo of Sacred Heart University proposed a different frame in 2008, the I.D.R. model. It names forces rather than phases: an initiating force, the longing for emotional reciprocation; driving forces, intrusive thoughts and fear of rejection; resultant forces, mood swings and impaired functioning. They placed limerence between obsessive-compulsive disorder and addiction while warning against collapsing it into either.

Neither model assigns durations. No stage has a published median length, nobody has timed the transition from crystallization to deterioration, and the 2026 paper closes by calling for formal diagnostic criteria, which is a polite way of reporting that none exist. Use the vocabulary to describe what you notice, not as a schedule.

When duration stops being the useful question

The instrument that matters here does not ask how long. The limerence-adapted Yale-Brown scale scores interference on five levels: none; slight interference with social or other activities, with overall performance unimpaired; definite interference with social or occupational performance, but still manageable; substantial impairment; and incapacitating. A separate item scores time occupied, with more than three hours a day landing in the upper bands.

Those anchors decide things a calendar cannot. Somebody at month three scoring "substantial impairment" is in more trouble than somebody at year six scoring "slight." Supporting figures point the same way: in the earlier experience-sampling subsample, 66% reported a mental health or neurodevelopmental diagnosis and 36% had already received treatment for limerence.

The strongest case against everything I have argued here is straightforward, and partly right. A person two weeks into an episode does not want a distribution. They want one number to hold, and every range you hand them invites them to pick the end matching their fear. I have watched that happen with yield ranges for fifteen years, which is why the halving-and-doubling ladder starts at a single figure and only then opens outward. Grant the objection its due: a range is worse than useless if the reader cannot carry it. It argues for a well-built centre, not for suppressing the spread.

One thing about these figures deserves the last word. Every duration number here, the two years, the six months, the four years, the 40-year maximum, came from people remembering backwards and estimating. In my trade that is a recall estimate, and recall estimates of duration run long. The only prospective measurement anyone has published ran seven days, which cannot observe the end of a two-year episode. Until somebody enrols a cohort at the start of an episode and follows it to the finish, the two-year figure carries an asterisk, and I would rather you knew where it was.

Frequently asked questions

How does limerence usually end?

Tennov described three exits. Consummation is reciprocation, which can convert the state into something better called love. Starvation is hope running out, leaving sadness and often relief. Transference moves the fixation to a new person. Her 1979 work was qualitative, and no study since has counted how often each ending occurs.

What are the stages of limerence?

Tennov's model has three: infatuation, crystallization, where the person is idealised and thoughts turn intrusive, and deterioration, where intensity fades. Wakin and Vo's 2008 model names forces instead: longing for reciprocation, then intrusive thoughts and fear of rejection, then mood swings and impaired functioning. Neither assigns durations to stages.

Can limerence become love?

By Tennov's consummation route, yes. When feeling is returned, limerence can settle into something she considered better described as love. Evans and colleagues define limerence by unreciprocated longing, so reciprocation ends the limerent state by definition. Neither claim rests on a study that followed couples forward over time.

How long does limerence last with no contact?

No study has measured it. Nobody has tracked a no-contact group against a contact group with a clock. Tennov argued that intermittent reciprocation sustains the state, which predicts that cutting contact shortens it. Note that 46.8% of limerent objects in one survey were friends or co-workers.

Can limerence last 10 years?

Yes. In the University of Sheffield survey of 235 people, the longest episode anyone reported ran 40 years, and the median longest episode was four years. Ten years sits comfortably inside the observed range, though far above the two-year central estimate from the 1,647-person study.

When does romantic preoccupation warrant professional support?

When it interferes, rather than when it lasts. The limerence-adapted Yale-Brown scale grades interference from none to incapacitating, and the level described as definite interference with work or social performance is the point clinicians act on. Preoccupation exceeding three hours daily also scores in the upper bands.