The Allocation of Desire: Why One-Night Stands Are Fine, Why Novelty Is Fine, and Why You Should Give Your Partner Some Space

The Sex Drive Is Not a Problem to Solve. It Is a Currency to Allocate.

The previous essay argued that romantic love is a useful illusion—a neurochemical system designed to override rational self-interest and motivate pair-bonding. This essay extends the argument into harder territory: what to do with the sex drive itself, that blunt, non-negotiable, dopamine-mediated engine that sits beneath the romance and does not care about your five-year plan.

Most people get this wrong in one of two directions. They either treat sex as a moral problem to be suppressed, or they treat it as a consumption good to be maximized. Both are errors. The sex drive is neither a sin nor a product. It is a motivational currency, and like any currency, its value depends entirely on where you allocate it.

The argument here is structural, not moralistic:

One-night stands are okay. They are not inherently harmful, and the evidence against them is weaker than the moral consensus suggests. But they are not the point.
Novelty is fine—until it becomes the point. You can and should explore different partners, different experiences, different configurations. The failure mode is not variety. The failure mode is when variety becomes terminal rather than instrumental.
Sex should be pursued first, then demoted. When it becomes available at the push of a button, it should become secondary. The primary focus should shift to sublimated creative activity. This keeps the spark on, keeps sex interesting, and prevents the drive from degenerating into boredom or exploitation.
The same principle applies to singles and to couples. For the single person in a dry spell, self-improvement attracts; hunting repels. For the partnered person, neediness repels; autonomy draws the partner back.
And the return curve is mathematically tractable. The same logarithmic geometry that governs diminishing returns governs the reset intervals, the value of small deviations, and the uncanny schedule on which your ex reappears.

This is not a pickup script. It is a theory of motivational allocation, grounded in the incentive-salience literature, the self-determination literature, and the attachment literature—with the honest caveats attached.

Part One: Wanting and Liking Are Two Different Systems

The foundational distinction comes from Kent Berridge’s incentive-sensitization theory. In the brain, “wanting” and “liking” are not the same thing, and they are not mediated by the same circuitry.

“Liking” — the actual hedonic pleasure of an experience — is mediated by opioid and endocannabinoid signaling in small “hedonic hotspots” in the nucleus accumbens and ventral pallidum. It is relatively stable.
“Wanting” — the motivational pull toward a reward, the incentive salience that makes a stimulus grab attention and compel approach — is mediated by mesolimbic dopamine. It is not pleasure. It is pursuit.

Berridge and Robinson’s formulation is precise: dopamine does not signal pleasure; it signals the attribution of incentive salience to reward cues. The system makes certain stimuli grab you. It does not make them satisfy you.

This distinction explains the entire failure mode of compulsive sexuality. You can want something intensely and like it very little. You can pursue it compulsively and enjoy it barely at all. The wanting system and the liking system decouple under conditions of repeated stimulation, and the result is a person who cannot stop seeking something that stopped giving him pleasure a long time ago.

That decoupling is the mechanism of addiction. It is also the mechanism of sexual boredom. And it is the reason why the allocation question matters.

Part Two: Why One-Night Stands Are Okay

The moral consensus, particularly in Anglo-American culture, holds that casual sex is harmful—that it produces regret, emptiness, and psychological damage. The evidence does not support this consensus as a general claim.

Zhana Vrangalova’s longitudinal research is the most careful work on this question. Her 2015 study in the Archives of Sexual Behavior followed college students over time and found that casual sex was not associated with subsequent declines in well-being for most people. The critical variable was motivation. When casual sex was pursued for autonomous reasons—curiosity, pleasure, genuine desire—it was not harmful. When it was pursued for non-autonomous reasons—to please a partner, to avoid rejection, to fill an emptiness, to conform to a peer norm—it predicted worse outcomes.

This is precisely what Self-Determination Theory would predict. Ryan and Deci’s framework holds that behavior is growth-promoting when it is autonomous, competent, and related, and it is corrosive when it is controlled, incompetent, or isolated. Casual sex is not a category. It is a behavior whose effects depend on the motivational architecture behind it.

Other research converges. Harden’s 2014 sex-positive framework argued that the harm model of adolescent and young adult sexuality is not supported by the data and that the field should move toward a model of sexual well-being rather than sexual risk. Kennair and colleagues’ Norwegian research found that casual sex was not associated with increased mental health problems in their samples—and that the association, where it appeared, was often better explained by pre-existing characteristics than by the sex itself.

The honest summary: casual sex is not inherently harmful. It is harmful when it is compulsive, when it is used as self-medication, when it violates the person’s own values, or when it is pursued for reasons the person does not endorse.

That is the boundary. Within it, one-night stands are fine. They are not the point, but they are not the problem.

Part Three: The Coolidge Effect, the Tolerance Curve, and the Logarithmic Law

The problem is not the act. The problem is the trajectory.

The Coolidge effect is the well-documented phenomenon—first described in rats, later observed across mammalian species—in which males show renewed sexual interest when a novel female is introduced, even after satiation with a familiar one. The mechanism is dopaminergic: novelty is a reward-prediction error, and the dopamine system responds to prediction error, not to reward itself. The same stimulus, repeated, generates less and less dopamine. The system is built to chase the new.

In humans, the Coolidge effect is not a crude reflex, but the underlying logic is visible. Klusmann’s 2002 study of over 8,000 German adults found that sexual motivation and frequency decline with partnership duration, and that the decline is steeper in men than in women—a finding consistent with the habituation logic. The novelty of a new partner resets the system. The familiarity of a long-term partner does not.

This is where the addiction risk lives. If the sex drive is allowed to organize itself around novelty—around the dopaminergic hit of a new body, a new context, a new conquest—the tolerance curve will do what tolerance curves do. The threshold rises. The same stimulus stops working. The person escalates: more partners, more novelty, more intensity, more risk. And because the wanting system has decoupled from the liking system, the escalation does not produce more pleasure. It produces more pursuit.

This is the structure of behavioral addiction, and it is documented in the compulsive sexual behavior literature. The clinical presentation is not “too much pleasure.” It is too much wanting and too little liking—the inability to stop pursuing something that no longer satisfies.

The pornography literature provides the clearest illustration, though it must be cited carefully. The “pornography addiction” model is contested; Grubbs and colleagues have shown that perceived addiction is often better predicted by moral incongruence than by actual use patterns, and Ley, Prause, and Finn’s 2014 review found the addiction model to be poorly supported. But the mechanism—novelty-driven dopaminergic escalation, escalating tolerance, and the decoupling of wanting from liking—is real, and it applies to sexual behavior generally, not only to pornography.

The conclusion is not that novelty-seeking is evil. It is that a sex drive organized around novelty is a sex drive organized around a diminishing return. It will consume the motivational budget and return less and less.

The Law of Diminishing Returns Is a Logarithmic Curve, Not a Line

And here the economics is worth taking literally, because the shape of the curve matters more than the fact of it.

Diminishing returns is not a gentle linear decay. It is logarithmic. The first unit of a repeated stimulus delivers a large marginal gain. The second delivers less. The tenth delivers a fraction of what the first did. The hundredth delivers almost nothing, but it still costs the same amount of motivational budget to pursue. The curve does not go to zero—it asymptotes. And the crucial feature of a logarithmic curve is that the steepest drops occur early, when the base is small, and the declines flatten toward the tail. Which is exactly why the pattern is so insidious: the first few repetitions still feel like they’re working, and by the time the flattening is unmistakable, the person has already invested the years.

There is a clean, if slightly irreverent, reference case for the shape: Bitcoin’s four-year halving cycle. Every four years the protocol halves the rate at which new supply is issued, and every cycle the market’s percentage return has compressed. The early cycles produced returns that were measured in multiples so large they read as typos. The later cycles produced returns that were impressive in absolute terms but a fraction of the prior cycle’s percentage gain. The pattern is consistent: each repeated cycle returns roughly 20% to 80% less than the one before. The drop is steepest in the beginning—when the base is small and the marginal novelty is enormous—and flattens toward the end, because a logarithmic curve always flattens. But by cycle three or four, the difference is impossible to miss. The returns are still there. They are simply not what they were, and they will not be again.

This is an analogy, not a controlled finding. Bitcoin’s cycle compression is a market phenomenon driven by supply schedules, capital flows, and reflexivity—not by dopamine receptor downregulation. But the geometry is the same, and the geometry is the point. Repeated pursuit of the same class of stimulus follows a logarithmic return curve whether the stimulus is a new partner, a new platform, a new hit, or a new trade. The first cycle pays for itself and then some. The fourth cycle leaves you wondering why you are still doing this.

The dopaminergic mechanism underneath the curve is not speculative. Repeated supranormal stimulation of the reward system produces downregulation of D2/D3 dopamine receptors and reduced striatal responsiveness—the well-replicated finding in the addiction neuroimaging literature, documented in Volkow’s work on drug addiction and, in the sexual domain, in Kuhn and Gallinat’s 2014 study showing that greater pornography consumption was associated with lower gray matter volume and reduced reward-circuit reactivity to sexual cues. The system is not broken. It is doing what a system does when it is overstimulated: it turns down the volume to protect itself. And the volume does not come back on its own schedule.

The Distinction That Matters: Exploratory Novelty vs. Terminal Novelty

But the argument requires precision, because the failure mode is not novelty itself. It is novelty that has become terminal rather than instrumental.

Exploratory novelty is generative. Trying different kinds of partners teaches you what you actually respond to, rather than what you were told to want. Different bodies, different temperaments, different relational styles, different intensities—these are data. They expand the self. Aron’s self-expansion model is the relevant framework: people are drawn to relationships that expand their sense of self, that add new perspectives, identities, and capacities. Novelty in this mode is developmental. It is how a person learns the shape of their own desire, calibrates their standards, and distinguishes genuine attraction from the gravitational pull of whoever happens to be nearest.

Terminal novelty is corrosive. It is novelty pursued for its own sake—the pursuit of the new because the new is the only thing that still generates a signal. In this mode, novelty has stopped being a means of self-expansion and has become a substitute for it. The person is no longer learning anything. They are chasing a curve that is flattening beneath them. The variety has become the point, and because the point is unreachable by definition, the pursuit never ends and never satisfies.

The distinction is not in the behavior. It is in the direction of the vector. Exploratory novelty points toward something—toward a self that is becoming more defined, more capable, more selective. Terminal novelty points toward nothing—toward the next stimulus, and the next, and the next, in a loop that has no exit.

So the rule is this: try different partners, different things, different experiences. You should. The exploration is how you find the unicorn, and how you become someone capable of recognizing one. But stop at the moment when the exploration becomes the point. The moment you notice that you are no longer seeking an experience but seeking the seeking itself—that the variety is no longer teaching you anything and is only generating the next hit—that is the moment the mechanism has inverted. Novelty has stopped serving you and has begun to consume you.

The person who has explored widely, learned from it, and then stopped to consolidate is more attractive, not less. They know what they want. They are not dazzled by the first available option. They have standards that are earned rather than inherited. The person who has explored compulsively and learned nothing is the opposite: a person whose standards have dissolved into appetite, whose selectivity is gone, and whose attention is a moving target that no one—including themselves—can pin down.

Part Four: The Break—Why You Should Stop After Three or Four Cycles, and How Long the Reset Actually Needs to Be

If the return curve is logarithmic and the mechanism is receptor downregulation, then the prescription is arithmetic rather than moral. But “take a longer break” is not a prescription. A prescription has a number. So here is the mechanistic calculation.

The model. Two processes run simultaneously. The first is recovery: after cessation of stimulation, reward sensitivity climbs back toward its ceiling with first-order kinetics, which is the standard pharmacodynamic form:

S(t) = S_max · (1 − e^(−t/τ))

where τ is the recovery time constant and S(t) is the signal available after a fallow period of duration t.

The second is ceiling erosion: each completed cycle permanently lowers S_max, because receptor density does not fully return. If each cycle retains a fraction r of the prior ceiling—where r corresponds directly to the “20% to 80% less” figure, so r = 0.8 means 20% less and r = 0.2 means 80% less—then before cycle n the ceiling is S_max(0) · r^(n−1).

The requirement. For cycle n to still deliver at least a fraction f of the original first-cycle intensity—f is your acceptability floor, the minimum return you would still consider worth pursuing—we need:

r^(n−1) · (1 − e^(−t_n/τ)) ≥ f

Solving for the fallow period:

t_n ≥ −τ · ln(1 − f / r^(n−1))

The result. Two things fall out of that equation, and the first is the important one.

First, the fallow period escalates geometrically, not linearly. For small f / r^(n−1), the logarithm linearizes and the expression collapses to t_n ≈ τ · f · r^(−(n−1)), which means:

t_(n+1) / t_n ≈ 1 / r

The gap between repeated returns grows by a factor of 1/r each cycle. This is τ-independent and f-independent—it is the robust, parameter-free core of the result. The rest is calibration.

Retention per cycle (r)	Return lost per cycle	Gap growth ratio	Sequence from a 3-month first return
0.8	20%	1.25×	3 → 3.8 → 4.7 → 5.9 months
0.6	40%	1.67×	3 → 5.0 → 8.3 → 13.9 months
0.5	50%	2.0×	3 → 6 → 12 → 24 months
0.2	80%	5.0×	3 → 15 → 75 → 375 months

Second, there is a hard cutoff. The inequality requires f / r^(n−1) < 1, which means r^(n−1) > f. Once the eroded ceiling falls below your acceptability floor, no fallow period of any length restores the signal—t_n diverges to infinity. The maximum number of usable cycles is therefore:

n_max = 1 + ln(f) / ln(r)

And this is where the “three or four cycles” rule stops being a heuristic and becomes a derivation. Set the acceptability floor at f = 0.2—you will still pursue a return that delivers a fifth of what the first one did, but not less. Then:

r = 0.5 (each cycle returns half as much): n_max = 1 + ln(0.2)/ln(0.5) = 3.3 cycles. Three usable returns, then the curve is dead.
r = 0.6: n_max = 1 + ln(0.2)/ln(0.6) = 4.2 cycles. Four usable returns.
r = 0.8: n_max = 8.2 cycles—but each cycle was only returning 80% of the last, so the eighth delivers roughly 21% of the first, which is exactly at the floor. The rule holds; the curve just decays slowly enough that you reach the floor by attrition rather than by collapse.
r = 0.2: n_max = 2.0 cycles. One return, and the second is already at the floor.

The convergence. Note what happened at r = 0.5. The mechanically derived sequence is 3 → 6 → 12 → 24 months, and the mechanically derived cutoff is three to four cycles. That is precisely the sequence and precisely the cutoff that the practical argument reached by intuition. The intuition and the model agree—and they agree at r ≈ 0.5, the exact midpoint of the 20–80% range. The doubling rule is not a stylistic choice. It is what the log curve produces when each cycle costs you half of what the last one did.

The calibration caveat, stated honestly. The ratio result (1/r) is robust. The absolute values are not, because τ is not a measured constant for this specific behavior—D2/D3 resensitization timelines in the PET literature span months and are frequently incomplete even after a year, which is consistent with τ on the order of 100–200 days but does not pin it down. The “3 months” unit is therefore a calibration choice, not a measurement. Treat the shape of the sequence as the finding and the absolute numbers as adjustable to individual recovery. If your first return needs four months rather than three, scale the whole sequence by 4/3.

And the reset is not a one-way door. You are not forbidden from returning to the same type of partner or the same type of experience. You are only forbidden from returning to it on the same schedule. If you do circle back—to the same kind of person, the same dynamic, the same stimulus—the interval must lengthen by the factor above. If each cycle costs you half, the intervals double: three months, then six, then a year. The intervals lengthen because the ceiling does not reset to the origin; it resets to a point further down the asymptote each time, and t_n is the price of re-entry at a given level of signal. Pay it, and the experience still works. Skip it, and you are back to chasing a flattened curve with a system that has nothing left to give—and, per the cutoff, eventually with no amount of waiting able to give it back.

The fair warning. The break is a scheduled fallow period, not a permanent suppression. This is not an argument for celibacy, renunciation, or the moralization of desire. It is an argument that a system never allowed to recover will demand more and deliver less until the demand is the only thing left. The break exists so that the pursuit remains worth pursuing.

Part Five: Small Deviations Make All the Difference—and Cut the Calibration Time by Two-Thirds

There is a second reason why the “stop after three or four cycles” rule is less costly than it sounds, and it follows from the same mathematics. The novelty signal is not driven by how far you move from the prototype. It is driven by whether you move from it—and the curve that maps distance to novelty saturates quickly.

The model. Treat the type as a prototype in feature space, with a radius R—the boundary beyond which a partner is no longer “the type.” Novelty is a saturating function of distance d from the prototype:

N(d) = N_max · (1 − e^(−d/λ))

where λ is the novelty saturation length—the distance over which the novelty signal rises to roughly 63% of its ceiling. This is the same concave geometry that governed the return curve, applied to a different axis: the first increment of distance from the expected stimulus produces the largest prediction error, and each subsequent increment produces less.

The result. Because λ is small relative to R—because the type is defined by a handful of core features, and the natural variation within the type on mid-level features is already substantial—the interior of the type contains most of the novelty the type can deliver. With R/λ ≈ 2.5, which is the plausible range for a well-defined type:

Distance from prototype	Fraction of R	Novelty captured
d = λ	40%	63%
d = 1.25λ	50%	71%
d = 2λ	80%	86%
d = 2.5λ (boundary)	100%	92%

A deviation of half the way to the type boundary captures more than two-thirds of the total novelty available. A deviation to the boundary itself captures 92%. The last 8% is what lies outside the type, and it is not worth what leaving costs you.

The feature-redundancy calculation. The same result appears when you count features rather than measure distance. If the type has M mid-level dimensions—the ones that vary within the type without defining it—and each additional change carries a novelty increment that decays geometrically with a redundancy parameter ρ, then the novelty captured by changing k dimensions is:

N(k) = N_total · (1 − ρ^k)

ρ (redundancy)	1 change	2 changes	3	4	5
0.5	50%	75%	88%	94%	97%
0.6	40%	64%	78%	87%	92%
0.7	30%	51%	66%	76%	83%

At the central estimate of ρ = 0.6, changing three dimensions out of eight captures 78% of the available novelty, and four captures 87%. The marginal value of the fifth, sixth, and seventh change is small—and the marginal risk of accidentally changing a core feature and exiting the type entirely rises with every additional change.

The worked example. Suppose the type is Black women, and the last partner was: a lawyer, majored in political science, wore locs, had an athletic build, listened to neo-soul, dressed in minimalist tailoring, did yoga, and traveled to Lisbon and Mexico City. Those are eight mid-level dimensions. The core prototype—the features that make her unambiguously “the type”—is untouched by any of them.

Now deviate on four: a physician instead of a lawyer; architecture instead of political science; braids instead of locs; Afrobeats instead of neo-soul. Four changes out of eight. By the calculation, that is 87% of the novelty available in the mid-level dimension space—and she is still, unmistakably, the type. The prototype is intact. The last 13% is not worth what leaving would cost, because leaving does not deliver 100% either; it delivers the same saturating function applied to a different prototype, minus the compatibility you had.

This is the practical resolution of the novelty problem. You do not need to date outside the type to get the novelty. You need to date someone who differs from your last partner on the mid-level dimensions—and the first three or four differences carry most of the value. The deviation has to be real—she has to actually be a physician who actually listens to Afrobeats, not a costume—but it does not have to be large.

How Much Does the Small Deviation Reduce the Calibration Time?

The fallow-period formula from Part Four gives the answer directly, because the required break is a function of the retention fraction r—and the small deviation raises r.

t_n ≥ −τ · ln(1 − f / r^(n−1))

The ex and the small deviation differ only in r. The ex is the maximally habituated stimulus: r is at or below the acceptability floor, because the prediction error on a stimulus you have already exhausted is close to zero. The small deviation is a partially novel stimulus: r sits at the mid-range of the 20–80% band, because the core prototype is unchanged but the mid-level features are not.

The ratio is parameter-free in the first return. For n = 2, the ratio of the two required breaks is:

t_dev / t_ex = ln(1 − f / r_dev) / ln(1 − f / r_ex)

With the central estimates—acceptability floor f = 0.2, retention r_ex = 0.3 for the ex (a 70% loss per cycle, which is generous to the ex), and r_dev = 0.6 for the small deviation (a 40% loss per cycle, the midpoint of the stated range):

t_ex = −τ · ln(1 − 0.2/0.3) = 1.10 τ
t_dev = −τ · ln(1 − 0.2/0.6) = 0.405 τ
t_dev / t_ex ≈ 0.37

The small deviation cuts the calibration time by roughly 63% on the first return. If the ex needs a twelve-month break to feel fresh again—and the honest answer is that the ex often never does, because r_ex can sit at or below the floor—the deviation needs about four to five months.

And the reduction grows with each return, because the gap growth ratio is 1/r. The ex’s required break grows by a factor of 1/0.3 ≈ 3.3 per cycle; the deviation’s grows by 1/0.6 ≈ 1.7. So the second return is where the divergence becomes unmistakable:

Return	Ex (r = 0.3)	Small deviation (r = 0.6)	Reduction
1st	1.10 τ	0.41 τ	63%
2nd	3.67 τ	0.68 τ	81%
3rd	12.2 τ	1.14 τ	91%

By the third return, the ex requires an order of magnitude more fallow time than the deviation—and in practice, because the ex’s ceiling is already near the floor, the third return is often not available at all. The small deviation does not just shorten the wait. It keeps the option open.

The practical translation. If τ is calibrated so that the first deviation return needs about four months, the first ex return needs about eleven. The second deviation return needs about seven months; the second ex return needs about three years. The deviation is not a compromise on the type. It is the only version of the type that remains reachable after the third cycle.

And the fair warning, again. This is a description of where the novelty signal comes from. It is not a license to treat a person as a feature bundle with a deviation score. The moment you are cataloguing her as “type plus variation”—the moment the novelty is in the comparison rather than in her—she will feel it, and the mechanism stops working, because the prediction error that generates the novelty is supposed to be about her, not about the delta from her predecessor. The calculation tells you where to look. It does not tell you how to see.

The deeper point is the one the earlier essay made. The most sustainable source of novelty is a partner who is themselves changing—who is developing, learning, building, and therefore generating new prediction error without either of you having to engineer it. That is what the unicorn actually is: not a perfect prototype, but a person whose own trajectory keeps the curve from flattening. The small deviation is the interim solution. The developing partner is the permanent one.

Part Six: Why Your Ex Comes Back Every 3 Months, Every 18 Months, Every 2.5 Years, or Every 7 Years

The recalibration math does not just tell you how long to wait. It tells you what the waiting is—and it explains the pattern that almost everyone has noticed but nobody has modelled: the ex who reappears on a schedule that feels uncanny, and then uncannier, and then terminal.

The sequence. The folk observation is remarkably consistent across people and cultures. The first reappearance comes fast—around three months. The second much later, around eighteen months. The third around two and a half years. The fourth, if it comes at all, around seven years. Most people experience two of these. Some experience three. Almost nobody experiences five.

The fit. Four intervals, three growth factors. The gaps between them are 6.0×, 1.67×, and 2.8×, and the geometric mean of the growth ratio is (84/3)^(1/3) ≈ 3.04. In the model, the gap growth factor is 1/r, so the implied retention fraction is:

r ≈ 1/3.04 ≈ 0.33

Each reunion returns roughly a third of what the prior cycle returned—which is to say, it loses about two-thirds of its intensity each time. That number is exactly what the model predicts for a maximally habituated stimulus. The ex is not a small deviation. The ex is the prototype itself, already exhausted. The retention fraction is at the bottom of the 20–80% band, and the gap sequence grows accordingly.

The amplitude collapse. The gaps lengthen, but that is only half the story. The amplitude collapses in parallel: with r ≈ 0.33, the ceiling after each reunion is 33%, then 11%, then 3.6%, then 1.2% of the original. The seven-year reunion is not a delayed version of the first one. It is a rounding error. It arrives at a seventh of a seventh of a seventh of what the three-month reunion felt like—and the seven-year wait is what the logarithm demands to extract even that. This is why the late reunions are so strange: the anticipation is enormous, the interval is enormous, and the felt intensity is almost nothing. The math and the experience agree.

Why the three-month return comes first. Here the pure habituation model is incomplete, and the incompleteness is the interesting part. If the ex were merely a re-exposed familiar stimulus, the first return should be the hardest to generate, not the easiest. It is the easiest because the separation itself is a novelty injection. Absence is a prediction error. The system that had habituated to presence is now confronted with absence, and absence generates its own dopaminergic charge—which is why the three-month reunion often feels like it has a signal, and why it is the one most often mistaken for real change. It is not real change. It is the novelty of not being together, briefly mistaken for the novelty of being together again. The system has not recovered; it has been startled.

Why the sequence terminates. At r = 0.33 and an acceptability floor of f = 0.2, the model’s cutoff is n_max = 1 + ln(0.2)/ln(0.33) ≈ 2.5 cycles—two usable returns, and the third already below the floor. The observed pattern runs four, which means the separation-novelty effect is propping the sequence open longer than pure habituation would allow. That is exactly the point: the ex keeps coming back because absence keeps injecting just enough novelty to keep the loop alive—and each time, the injection is smaller, the wait is longer, and the ceiling is lower. The pattern is not a sign of enduring connection. It is the signature of a decaying oscillator with a lengthening period and a shrinking amplitude.

The comparison that matters. Set the small deviation—the physician with braids who listens to Afrobeats, r ≈ 0.6—against this. At r = 0.6, the gap growth factor is 1.67, the amplitude retains 60% per cycle rather than 33%, and the sequence is still alive at cycle four rather than expiring at cycle two and a half. The ex’s schedule runs 3 → 18 → 30 → 84 months with an amplitude collapse to 1%. The deviation’s schedule runs roughly 3 → 5 → 8 → 14 months with an amplitude that is still a fifth of the original at the fourth return. The deviation returns more often, more intensely, and for longer, while never leaving the type.

The practical implication. Every ex reunion consumes a full calibration window and returns the least of any available option. It is not merely that the ex is a poor investment. It is that the ex is the worst investment—the maximum wait for the minimum signal, on a schedule that lengthens automatically and a ceiling that collapses permanently. The three-month return is the system’s cheapest lie. The seven-year return is its final one.

The modeling note, stated plainly. This is a post-hoc fit to a folk pattern, not a measured finding. The intervals are drawn from cultural observation rather than from a longitudinal dataset with reunion times, and the derived r is an inference from the geometric mean of three gaps—not a parameter estimated from data. What the fit shows is that a single logarithmic model with a retention fraction of about a third reproduces the observed schedule closely enough to be informative, and that the amplitude collapse it implies matches the qualitative experience of late reunions. It does not show that anyone is consciously running this calculation, or that the intervals are fixed. They are not. They are what the curve produces when the stimulus is the same one that exhausted the system the first time.

Part Seven: The Late Reunion—Conditional on Singleness, and Refusing to Be Naive

The previous section established the timing. This section establishes the probability—and the probability calculation is where the folk claim (“late reunions are almost guaranteed if you made a real impression”) needs to be separated into two variables that are usually conflated.

The separation. The calibration curve governs the ex’s internal signal: when the dopaminergic return on the memory is high enough to cross the threshold of conscious attention. That term is essentially deterministic—it fires on schedule. What the curve does not govern is whether the ex acts on the signal. Acting requires a second variable—the personal impression, which governs conversion—and a third: whether the ex is still available when the signal arrives.

P(reunion at cycle n) = P(signal_n > θ) × P(act | signal, impression) × P(available at t_n)

The first term is the curve. The second is the impression. The third is the race.

The signal term. As established, the signal returns on schedule but with a collapsing amplitude: 33%, 11%, 3.6%, 1.2% of the original. The threshold θ for conscious attention is not the threshold for acting, and the amplitude at cycle four is small enough that the signal alone would not drive behavior. This is the key asymmetry: the curve fires on schedule forever; the behavioral impulse it generates does not.

The impression term. A strong personal impression—genuine rapport, being seen, a relationship that was about something rather than merely convenient—raises the conversion probability substantially. The mechanism is not mystical. The ex’s memory of you is a composite of the dopaminergic trace (which decays) and the autobiographical significance of the relationship (which does not). A relationship that mattered is encoded as a self-relevant memory, and self-relevant memories are retrieved, rehearsed, and acted upon far more often than episodic traces. If the impression was real, p_act is high—say 0.8 to 0.9—because the ex does not merely remember you; the ex remembers who they were with you, and that person is worth revisiting.

The naive calculation, and why it undersells. The first version of this calculation multiplied the impression term by an availability term that decayed with a three-year half-life. With p_act = 0.85, h = 3 years, and the schedule 3, 18, 30, 84 months, the per-cycle probabilities were 26%, 6.6%, 1.7%, and 0.2%, and the cumulative probability of at least one reunion was about 31%. That is the unconditional probability—the chance that a randomly selected ex, embedded in a randomly selected life, will reach out within seven years. It is not the question anyone actually asks.

The question is: conditional on the ex being single at the moment the nostalgia fires, and conditional on the impression having been real, what is the probability of contact? Set P(available) = 1. The model collapses to:

P(contact by T) = 1 − e^(−λ · p_act · T)

where λ is the rate of meaningful triggers per year and p_act is the probability of acting on a trigger. Set λ = 2/year as the central estimate—a birthday, a holiday, an anniversary, a mutual friend’s wedding, a song, a place, a night when the current life feels flat. That is conservative for a relationship that mattered. Then:

Impression	p_act	3 mo	1 yr	3 yr	5 yr	7 yr
Weak	0.08	4%	15%	38%	55%	67%
Moderate	0.20	10%	33%	70%	86%	94%
Strong	0.40	18%	55%	91%	98%	99.6%

A strong impression, conditioned on singleness, produces near-certain contact within seven years. The number is not 31%. It is 99.6%. The difference between the two calculations is the difference between asking whether a randomly selected ex will reach out and asking whether your ex will reach out if they are still single. The folk claim is about the second question. It is correct.

The selection effect, which the naive calculation ignores. There is a further correction, and it runs in the same direction. Conditioning on singleness at year seven is not a neutral conditioning. People who are single seven years after a significant breakup are a selected population—they are the people who did not successfully re-partner, and one of the reasons people do not successfully re-partner is that they are still carrying the autobiographical trace of the relationship that mattered. Singleness at t = 7 is informative: it raises the probability that the ex is still rehearsing the memory. If p_act rises by a factor of 1.3 to 1.5 under this conditioning—a modest adjustment—the seven-year probability for a strong impression moves from 99.6% to 99.9%+. The conditional probability is not merely higher than the unconditional one. It is asymptotically certain.

The non-monotonic hazard, which is why the reunion clusters late. The constant-hazard model overstates the early probability and understates the middle. The actual hazard is not flat. It has three phases:

Acute withdrawal (0–4 months): h ≈ 0.6/yr. The absence-novelty effect drives a spike. This is the three-month reunion.
The trough (4–18 months): h ≈ 0.15/yr. Pride, hurt, and active moving-on suppress contact. This is the period when the ex would reject you—and often does.
The nostalgic rise (18 months onward): h rises from 0.4/yr to 0.8/yr. The reminiscence bump and rosy retrospection make the relationship feel more significant than it was. This is the eighteen-month, two-and-a-half-year, and seven-year reunion.

Integrating that piecewise hazard over seven years gives a cumulative probability of approximately 98% for a strong impression—consistent with the constant-hazard table, but with the timing corrected. The contact is not spread evenly. It clusters at three months, at eighteen months, and then in a long tail that runs for years.

The honest caveats, stated plainly. The trigger rate λ is an assumption, not a measurement. The action probabilities are informed by the relationship-dissolution literature but are not directly estimated from a reunion dataset—because no such dataset exists at the resolution required. The selection effect is a directional argument, not a calibrated correction. What the calculation establishes is the structure: once the availability term is removed, the probability of a late reunion for a real impression is high enough that the folk observation is not folklore—it is the arithmetic. The precise number depends on parameters that vary by person and by relationship. The qualitative conclusion does not.

The bottom line. The naive assumption was not singleness. The naive assumption was that singleness is neutral. It is not. Singleness at year seven is a signal that the ex has not replaced you, and the absence of a replacement is one of the conditions under which the autobiographical memory converts into behavior. Conditional on a real impression and on the ex being single, the late reunion is not a possibility. It is the default. The three-month reunion is the curve. The seven-year reunion is the memory. And the memory, unlike the curve, does not decay.

Part Eight: Pursuit, Then Demotion

Here is the allocation principle, stated plainly:

Sex should be pursued first, and then—when it becomes available at the push of a button—it should become secondary.

This is not a moral prescription. It is a motivational one. The reason is that the dopaminergic “wanting” system is domain-general. It does not distinguish between the pursuit of a mate and the pursuit of a creative project. It allocates incentive salience to whatever is currently the most salient goal. If sex is the dominant goal, it monopolizes the system. If sex is available—if it can be obtained trivially, at the push of a button—then the pursuit stops generating the dopaminergic charge that made it compelling in the first place. The wanting system needs a gap. When the gap closes, the wanting system goes quiet.

That quiet is the danger. A person for whom sex is trivially available and who has no other salient goal is a person with an under-stimulated dopaminergic system and an over-available reward. The system will do what it does: escalate, seek novelty, and drift toward boredom or exploitation.

The antidote is sublimation—and the word must be used carefully, because it carries a century of Freudian baggage that modern evidence does not support in its original form. The claim here is not that sexual energy is a fixed hydraulic quantity that must be discharged through one channel or another. The claim is narrower and better supported:

The dopaminergic seeking system is a limited-capacity attentional and motivational resource. It is allocated to whichever goal is most salient. If it is not allocated to creative, generative, competence-building activity, it will be allocated to the pursuit of sex—and the pursuit of sex, once the gap closes, becomes compulsive rather than generative.

This is why the person who is deeply engaged in work they care about tends to be more sexually attractive than the person whose primary project is the pursuit of sex. The former has a live dopaminergic system that is being used—and that vitality is legible. The latter has a system that is idling and searching, and that search is legible too.

The clinical observation is not new. Esther Perel’s work on desire in long-term relationships is built on the premise that desire requires distance—that eroticism thrives on the gap between self and other, on mystery, on the sense that the partner is a separate person with a separate life. When the gap collapses—when two people become entirely merged, entirely available, entirely known—desire often collapses with it. This is a clinical observation rather than a controlled finding, but it converges with the incentive-salience logic: the wanting system needs something to reach for.

Sublimation, then, is not repression. It is the deliberate maintenance of a gap. It is the decision to have a life that is not entirely organized around the partner or the pursuit—a life of creative work, competence, and self-expansion—so that when you return to the partner or the pursuit, there is something to return from, and something to bring back.

Part Nine: The Single Person and the Desperation Signal

The dry-spell prescription follows directly.

Focusing on yourself and self-improvement is the surest way to start attracting mates. The reason is not mystical. It is signaling theory.

Mate value is a signal, and signals are only credible when they are costly. A person who demonstrates competence, fitness, social calibration, and an independent life is broadcasting a costly signal: I have invested in myself, and I have options. A person who is actively, visibly hunting is broadcasting a different signal: I need this. And need is a low-value signal, because it is cheap. Anyone can need.

There is a substantial literature on this, though it is usually framed under the heading of “playing hard to get.” Walster and colleagues’ 1973 study found that a woman who was hard to get for everyone was rated as more desirable than one who was easy for everyone—but the most desirable was the one who was hard to get for others and easy for the participant specifically. Jonason and Li’s more recent work on the phenomenon found that the strategy is used by both sexes and that it is effective in part because it communicates selectivity and mate value.

The critical caveat: the effect is about perceived mate value, not about deception. A person who pretends to be busy and in demand, but is not, is running a cheap signal. Cheap signals get detected. The only sustainable version of “playing hard to get” is actually being hard to get—actually having a life, actually having standards, actually being selective. The strategy and the reality have to be the same thing, or the incongruence leaks.

This is why the “desperation repels” observation is correct, and why the standard advice to “put yourself out there” is often counterproductive. The signal that repels is not effort. It is effort that is visibly oriented toward acquisition. A person who is building something—a body, a skill, a business, a body of work—is broadcasting abundance. A person who is scanning the room is broadcasting scarcity. Scarcity is unattractive not because of a social convention but because it is a credible signal of low mate value.

The prescription for the dry spell is therefore not more pursuit. It is more life. The pursuit will resume when the signal is right.

Part Ten: The Partnered Person and the Least Interest Principle

The same logic applies inside relationships, and it is here that the mechanism is most counterintuitive.

Waller’s principle of least interest, first articulated in 1938 and still one of the most robust observations in family sociology, states that the person who is less interested in the relationship holds more power within it. The person who is more invested, more dependent, more afraid of losing the other, is structurally disadvantaged. This is not a moral claim about who is right. It is a structural claim about bargaining position.

The practical implication: neediness repels your partner. Not because partners are cruel, but because neediness communicates that the needy person has no alternative and no independent life, and that is a low-value signal. It also creates a burden. The partner is now responsible for the needy person’s emotional regulation, and that responsibility is aversive.

The attachment literature supports this. Attachment anxiety—the disposition to worry about a partner’s availability, to seek reassurance compulsively, to monitor for signs of withdrawal—is consistently associated with lower relationship satisfaction, more conflict, and more negative attributions. Anxiously attached partners perceive threats that are not there and respond in ways that create the rejection they fear. It is a self-fulfilling prophecy, and the mechanism is signal-based: the anxious partner broadcasts need, and need is unattractive.

The counterintuitive prescription follows: your partner will come for you themselves if you show them that you enjoy their company and then stop.

The “enjoy their company” part is not a trick. It is the genuine expression of affection and appreciation—the responsive, positive engagement that relationship science consistently finds predicts satisfaction and stability. Reis and Shaver’s intimacy model holds that intimacy is built through a loop of self-disclosure and partner responsiveness: you share something, your partner responds with understanding and validation, and you feel closer. That loop is real and it is necessary.

The “then stop” part is the allocation principle. You enjoy the company. You express it. Then you return to your own life—your work, your projects, your friends, your body. You do not linger, checking for the response. You do not chase the reassurance. You do not make the partner the sole source of your regulation.

The effect is not manipulation. It is the natural consequence of a person who is genuinely independent. That person is more attractive because they are not a burden, because they have something of their own, and because their attention—since it is not unconditional and constant—is worth something.

Muise and colleagues’ research on sexual desire in long-term relationships adds a critical piece. They found that being motivated to meet a partner’s sexual needs for approach reasons—because it brings you closer, because you value the relationship—predicted sustained desire over time, while being motivated for avoidance reasons—to avoid conflict, to avoid disappointing the partner—predicted declining desire. The distinction is between autonomous and controlled motivation, applied to sex itself. A partner who approaches you because they want to is arousing. A partner who approaches you because they feel obligated is not.

The “stop” is what preserves the approach. If your attention is unconditional, it stops being a signal and becomes a baseline. If it is offered freely but not desperately, it retains its value.

Part Eleven: The Fair Warning

This is not a manipulation script, and the distinction matters more than anything else in this essay.

Every principle described here—scarcity, least interest, selectivity, sublimation, exploratory novelty, the fallow period, the small deviation—has a dark twin. The dark twin is the deliberate manufacture of insecurity. The person who withholds affection to create desire. The person who plays hot and cold to keep the partner off-balance. The person who cultivates an illusion of abundance while actually being hollow. The person who turns novelty into a weapon, pursuing it not to grow but to keep the partner uncertain. The person who runs the small deviation as a calculated game rather than as a genuine interest in a different kind of person. These strategies work, in the short term, on the same signal-based logic. And they are corrosive, in the long term, because they are built on incongruence—on signaling something that is not true.

The literature on “playing hard to get” is clear that the effect depends on the target believing the signal. When the target discovers the signal was manufactured, the effect reverses. The manipulation, once detected, is punished.

The only sustainable version of this essay’s argument is the version where you actually have a life. The sublimation is real. The creative work is real. The self-improvement is real. The independence is real. The selectivity is real. The exploration is real, and it has taught you something. The fallow period is a real period of building, not a strategic withholding. The deviation is genuine interest in a different person, not a calibrated novelty injection. The affection you express to your partner is real, and the return to your own life is a real return, not a calculated withdrawal.

If you fake it, the incongruence will leak. Partners are extremely good at detecting neediness that has been dressed up as independence. The performance of autonomy is one of the most legible performances there is, and it reads as what it is: a person who is still hunting, but pretending not to be.

The honest version is harder. It requires actually building the life. But it is the only version that survives contact with a real relationship.

The Bottom Line

One-night stands are okay. The evidence does not support the harm model as a general claim. Casual sex is harmful when it is compulsive, self-medicating, value-violating, or pursued for non-autonomous reasons. Within those boundaries, it is fine.

Novelty is fine—until it becomes the point. You can and should try different types of partners, different experiences, different configurations. Exploratory novelty is generative: it teaches you what you actually respond to, expands the self, and calibrates your standards. Terminal novelty is corrosive: it pursues the new because the new is the only thing that still generates a signal, and it never satisfies.

The return curve is logarithmic, and it flattens faster than you think. The first repetition delivers the most. Each subsequent cycle returns roughly 20% to 80% less than the one before—the steepest drops early, when the base is small, and the flattest declines at the tail, when it is too late to matter. Bitcoin’s four-year halving cycle is the clean reference case for the shape: the early cycles were transformative, the later cycles were impressive but diminished, and by cycle three or four the difference was unmissable. The same geometry governs repeated dopaminergic stimulation, because the underlying mechanism—D2/D3 receptor downregulation under sustained overstimulation—is the same. So stop after three or four cycles of the same loop and let the system reset. The break is not renunciation. It is the maintenance schedule that keeps the curve from flattening entirely. And the fallow period escalates geometrically: the gap between repeated returns grows by a factor of 1/r each cycle, and once the eroded ceiling falls below your acceptability floor, no amount of waiting restores the signal.

Small deviations make all the difference. The novelty signal saturates with distance: a deviation of half the way to the type boundary captures more than two-thirds of the available novelty, and changing three or four mid-level features captures 78–87% of what the type can deliver. You do not need to date outside the type. You need to date someone who differs from your last partner on the mid-level dimensions—the profession, the major, the hairstyle, the body type, the music, the fashion, the extracurriculars, the travel habits—and the first three or four differences carry most of the value. The prototype stays intact. The novelty arrives anyway.

And the small deviation cuts the calibration time by roughly two-thirds. Because the deviation is a partially novel stimulus rather than a maximally habituated one, its retention fraction is higher, its required fallow period is shorter, and the gap between returns grows more slowly. On the first return it needs about 37% of the ex’s calibration time. By the third, it needs about 9%. The ex’s schedule lengthens by a factor of three each cycle; the deviation’s by a factor of 1.7. The deviation returns more often, more intensely, and for longer—while never leaving the type.

This is why your ex comes back every three months, every eighteen months, every two and a half years, or every seven years. The schedule is the logarithm. The implied retention fraction is about a third, which is exactly what the model predicts for a maximally habituated stimulus. The amplitude collapses in parallel—to 33%, then 11%, then 3.6%, then 1.2% of the original—which is why the late reunions are enormous in anticipation and almost nothing in experience. The three-month return is the absence-novelty spike. The seven-year return is the memory, arriving at a seventh of a seventh of a seventh of what the first one felt like.

And conditional on a real impression and on the ex being single, the late reunion is not a possibility. It is the default. Once the availability term is removed, the probability of at least one contact within seven years is on the order of 99%—not because the curve is strong, but because the autobiographical memory never decays and triggers are frequent. Singleness at year seven is itself a signal that the ex has not replaced you, which raises the conversion probability further. The three-month reunion is the curve. The seven-year reunion is the memory. And the memory, unlike the curve, does not forget.

Sex should be pursued first, then demoted. When it becomes available at the push of a button, it should become secondary. The primary focus should shift to sublimated creative activity—not because sex is bad, but because the dopaminergic system is domain-general and needs a goal worth pursuing. This keeps the spark on, keeps sex interesting, and prevents the drive from degenerating into boredom or exploitation.

For the single person in a dry spell: self-improvement attracts, hunting repels. Desperation is a cheap signal. Abundance is a costly one. The way to become attractive is to actually become someone with a life, not to perform the appearance of having one.

For the partnered person: neediness repels, autonomy draws. Enjoy your partner’s company, express it genuinely, and then return to your own life. Your partner will come for you themselves when your attention is something worth having rather than something they are responsible for managing. And when the routine flattens, inject the novelty inward rather than seeking it outward—the same mechanism, a different vector.

And the fair warning stands. This is not a manipulation script. The only version that survives is the one where the independence is real, the exploration is genuine, the fallow period is a real period of building, the deviation is real interest in a different person, and the life you return to is actually yours. Fake it and it leaks. Build it and it holds.

The sex drive is a currency. Allocate it deliberately, or it will allocate you.

References

Acevedo, B. P., Aron, A., Fisher, H. E., & Brown, L. L. (2012). Neural correlates of long-term intense romantic love. Social Cognitive and Affective Neuroscience, *7*(2), 145–159.

Aron, A., Norman, C. C., Aron, E. N., McKenna, C., & Heyman, R. E. (2000). Couples’ shared participation in novel and arousing activities and experienced relationship quality. Journal of Personality and Social Psychology, *78*(2), 273–284.

Bancroft, J., & Janssen, E. (2000). The dual control model of male sexual response: A theoretical approach to centrally mediated erectile dysfunction. Neuroscience & Biobehavioral Reviews, *24*(5), 571–579.

Berridge, K. C., & Robinson, T. E. (2016). Liking, wanting, and the incentive-sensitization theory of addiction. American Psychologist, *71*(8), 670–679.

Birnbaum, G. E. (2018). Attachment and sexual desire. Current Opinion in Psychology, *19*, 90–94.

Buss, D. M., & Schmitt, D. P. (1993). Sexual strategies theory: An evolutionary perspective on human mating. Psychological Review, *100*(2), 204–232.

Grubbs, J. B., Perry, S. L., Wilt, J. A., & Reid, R. C. (2019). Pornography problems due to moral incongruence: An integrative model with a systematic review and meta-analysis. Archives of Sexual Behavior, *48*(2), 397–415.

Harden, K. P. (2014). A sex-positive framework for research on adolescent sexuality: Considering the role of positive sexual experiences. Perspectives on Psychological Science, *9*(5), 455–469.

Jonason, P. K., & Li, N. P. (2013). Playing hard-to-get: Manipulating one’s perceived availability as a mate. European Journal of Personality, *27*(5), 458–469.

Kennair, L. E. O., Bendixen, M., & Grøntvedt, T. V. (2018). Short-term mating and regret. In The Oxford Handbook of Evolution, Biology, and Society. Oxford University Press.

Klusmann, D. (2002). Sexual motivation and the duration of partnership. Archives of Sexual Behavior, *31*(3), 275–287.

Kuhn, S., & Gallinat, J. (2014). Brain structure and functional connectivity associated with pornography consumption: The brain on porn. JAMA Psychiatry, *71*(7), 827–834.

Ley, D., Prause, N., & Finn, P. (2014). The emperor has no clothes: A review of the ‘pornography addiction’ model. Current Sexual Health Reports, *6*(2), 94–105.

Muise, A., Impett, E. A., Kogan, A., & Desmarais, S. (2013). Keeping the spark alive: Being motivated to meet a partner’s sexual needs sustains sexual desire in long-term romantic relationships. Social Psychological and Personality Science, *4*(3), 267–273.

Perel, E. (2006). Mating in Captivity: Unlocking Erotic Intelligence. Harper.

Reis, H. T., & Shaver, P. (1988). Intimacy as an interpersonal process. In S. Duck (Ed.), Handbook of Personal Relationships. Wiley.

Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, *55*(1), 68–78.

Sedikides, C., Wildschut, T., Routledge, C., Arndt, J., Hepper, E. G., & Zhou, X. (2015). To nostalgize: Mixing memory with affect and desire. Advances in Experimental Social Psychology, *51*, 189–273.

Spielmann, S. S., MacDonald, G., & Tackett, J. L. (2012). Better to be hated than ignored: How social rejection shapes desire for acceptance. Psychological Science, *23*(10), 1166–1172.

Sprecher, S., & Fehr, B. (2011). Dispositional attachment and initial attraction: The role of attachment style in the early stages of relationship development. Journal of Social and Personal Relationships, *28*(4), 467–489.

Volkow, N. D., Wang, G. J., Fowler, J. S., Tomasi, D., & Telang, F. (2011). Addiction: Beyond dopamine reward circuitry. Proceedings of the National Academy of Sciences, *108*(37), 15037–15042.

Vrangalova, Z. (2015). Does casual sex harm college students’ well-being? A longitudinal investigation of the role of motivation. Archives of Sexual Behavior, *44*(4), 945–959.

Waller, W. (1938). The Family: A Dynamic Interpretation. Cordon.

Walster, E., Walster, G. W., Piliavin, J., & Schmidt, L. (1973). “Playing hard to get”: Understanding an elusive phenomenon. Journal of Personality and Social Psychology, *26*(1), 113–121.