Search the question and a number comes back within seconds: testosterone rises about 45 percent on day seven of abstinence. It is on video thumbnails, in forum sidebars, on supplement pages and in app onboarding screens. It comes from one paper. That paper enrolled 28 men, measured no mood, energy or performance outcome of any kind, and was withdrawn by its journal in 2021. That withdrawal is the least known important fact in this subject, and it is where an honest answer has to start.
What this article is assessing
The practice is much older than the hormone claim, and it comes from traditions that were not asking the hormone question. Brahmacharya, in Hindu and Jain thought, is a discipline of conduct, framed in its own texts as a matter of vows, restraint and the ordering of a life. Daoist teaching on the conservation of jing belongs to a cosmology of vital forces with its own account of what is conserved and to what end. Ayurvedic accounts of shukra and ojas use that tradition's own vocabulary of vitality. Each is a coherent system with its own aims, and none of them is making a claim about serum testosterone. Their cosmologies are not physiology, and neither grading them by hormone assay nor recruiting them as corroboration would be honest. This article does not attempt to assess them.
Four things also get used interchangeably and are not the same. Celibacy is abstaining from sexual relationships. Continence is a broader discipline of restraint, not necessarily limited to sex. Retention is abstaining from ejaculation, which does not by itself mean abstaining from sex. Non-ejaculatory technique is a set of practices for sexual activity or orgasm without ejaculation, which is different again. The studies below tested none of these as a practice. They tested fixed windows of abstinence from ejaculation in laboratory volunteers.
So the question here is narrow. Does abstaining from ejaculation raise testosterone, and is raised testosterone the reason men report feeling different?
The study everyone cites, and what happened to it
Abstaining for seven days raises your testosterone by about 45 percent.
The figure traces to one research group. Jiang and colleagues measured serum testosterone daily in 28 male volunteers during abstinence following ejaculation, reporting minimal fluctuation on days 2 to 5 and a peak on day 7 at 145.7 percent of baseline (P<0.01). The English-language version everyone cites was retracted in 2021. It had no control condition and measured no behavioural outcome, and in more than twenty years it has never been independently replicated.
The steelman is real. This was a daily measurement rather than a single endpoint, which is a better design than most of what circulates in this area, and it is biologically plausible that the hypothalamic-pituitary-gonadal axis responds to ejaculation on some timescale. Nobody should pretend the hypothesis is stupid.
The problems, in order of weight. The English-language version was retracted by the journal's Editor-in-Chief in 2021 because it substantially overlapped with the group's own earlier Chinese-language publication; the retraction note records that two of the four listed authors could not be contacted. Twenty-eight men is a small sample for a hormone that varies substantially within a single man across a single day and between days. There was no control group. And decisively for anyone hoping to act on the result, the study measured no energy, mood, confidence, cognition or performance outcome at all.
The most quoted number in semen retention comes from a paper its own journal has withdrawn.
What the retraction does and does not mean
It was a retraction for duplicate publication, not for fabricated data. The accurate summary is not that the study was fraudulent, and anyone reaching for the word debunked is overreaching in the other direction. The original Chinese-language report of the same 28 men exists and carries the same day 7 figure. That is the same dataset, not an independent replication of it. We have not checked its current standing in any retraction database, so we make no claim about that either way.
This sets a discipline the rest of the article has to keep. The paper is an object of study in this discourse, not a source of evidence in it. It cannot be used to argue that retention raises testosterone. It equally cannot be mined for a convenient negative finding about what happens after day seven, and no later section here borrows support from it or agreement with it. Set aside means set aside.
The only study that went past a week enrolled ten men
Three weeks of abstinence raises testosterone.
Exton and colleagues studied 10 healthy adult men, each of whom completed a laboratory session of arousal and masturbation-induced orgasm, abstained for three weeks, then repeated the session. Orgasm itself did not alter plasma testosterone in either session. Higher testosterone concentrations were observed after the abstinence period. Ten men, no control group, and a fixed rather than counterbalanced order. Not replicated in twenty-five years.
The strength is real: a prospective within-subject human design, direct plasma measurement, and the longest abstinence window anyone has used. It was run by a group independent of the Jiang team, which matters mainly for what it is not. It is not a replication of the withdrawn paper, and it cannot draw strength from pointing in the same direction as one.
The weakness is structural. Every participant did the baseline session first and the abstinence session second. Nothing was counterbalanced. Any drift across three weeks from seasonality, sleep, training load, familiarity with an unusual laboratory procedure, or plain regression to the mean is fully confounded with the abstinence effect. A waitlist arm or a reversed order would have cost very little and was not run. Nothing was sampled during the three weeks either, so the study says nothing about the shape of any trajectory. Secondary sources also disagree about whether the testosterone difference showed up at resting baseline or only during the arousal phase, and we could not resolve that from the published abstract. Those are two different claims, and the distinction is unverified.
Does it keep climbing?
Testosterone keeps rising the longer you go, so a 90 day streak beats a 30 day streak.
No adequately powered longitudinal study has tracked serial morning testosterone in men across abstinence periods of a month or more. Not one. The charts showing a line rising through week two, week four and week twelve have no dataset behind them: the continuation is drawn, not measured. Anyone quoting a figure for day 30, day 60 or day 90 is not reporting a finding, because no finding exists.
Nobody has looked is a different answer from it has been disproven, and the difference matters. A practitioner who says the science has not run this study is being more accurate than either side of the usual argument.
That said, for a claim this heavily marketed and this cheap to test, twenty years without a single properly powered longitudinal study is itself informative about how much of this field is measurement and how much is repetition.
The abstinence claim rests on Jiang and Exton. Neither reached thirty men. The British national probability sample discussed below, at 1,599 men, is off this scale entirely.
See the numbers
| Isenmann 2021 (completers) | 8men |
|---|---|
| Exton 2001 (3 weeks) | 10men |
| Jiang 2003 (retracted) | 28men |
| Catena 2024 (daily diary) | 41men |
| Testosterone Trials 2016 | 790men |
The claim that runs the other way
Ejaculating crashes your testosterone.
Several studies have looked for the drop using different methods and none has found it. Exton measured plasma testosterone directly through masturbation-induced orgasm in 10 men and found it unaltered. Isenmann ran a randomised three-arm single-blind crossover, 11 enrolled and 8 completing, and found that both masturbation with a visual stimulus and the visual stimulus alone appeared to counteract the normal diurnal decline in free testosterone. In a British national probability sample of 1,599 men using morning saliva and mass spectrometry, testosterone was positively associated with recent and frequent masturbation: men who had masturbated in the past seven days had higher levels than men who had not masturbated in over a year.
The limits are substantial. Eight completers is a pilot, and it describes hours rather than days. Dabbs, which found salivary testosterone rising across evenings on which intercourse occurred, studied four couples. The national sample is cross-sectional, so causation could run the other way, with higher testosterone driving more sexual activity rather than the reverse, and it does not test abstinence at all. None of these followed men for weeks.
Not one of them shows the drop the claim requires, and two point in the opposite direction. Post-ejaculation flatness is real as an experience and men are not imagining it. Whatever produces that state, the measurements that exist say it is not falling testosterone.
The step almost nobody takes
Suppose abstinence does move testosterone. A second question decides whether that would matter, and almost nobody asks it: within the normal range, does a testosterone number predict energy, drive and confidence?
The best available answer comes from the Testosterone Trials, which randomised 790 men aged 65 and over with serum testosterone below 275 ng/dL to a year of testosterone gel or placebo. Attribution matters here, because it is routinely mangled. The 2016 New England Journal of Medicine paper reports three of the trials: Sexual Function, Physical Function and Vitality. The Cognitive Function and Cardiovascular trials were published separately and are not in that paper, and nothing below is drawn from them.
Both directions, then. The Vitality Trial, 474 men selected for self-reported low vitality, missed its primary outcome: testosterone did not increase the proportion of men achieving a meaningful improvement on the fatigue measure. The Physical Function Trial, 387 men, also missed its primary endpoint, although the walking-distance improvement was significant among all participants. The same paper reports that testosterone slightly improved vitality, mood and depressive symptoms on the continuous measures. And the Sexual Function Trial, 459 men, met its primary outcome, with moderate increases in sexual activity and desire.
Correcting low testosterone brings energy and vitality back.
In 790 men aged 65 and over with clinically low testosterone, a year of testosterone gel missed the primary vitality endpoint in the 474 men of the Vitality Trial, while producing small improvements in vitality, mood and depressive symptoms on the continuous measures. Sexual function did improve on its primary outcome in 459 men. So something moved, mostly at the low end and mostly in the sexual domain, and the headline energy result was the one that failed. Confidence was not an outcome in any of these trials. These were men whose levels were clinically low, not men within the normal range.
Within the normal range
Most practitioners are not hypogonadal, so the within-range evidence is the evidence that applies to them. In the same national sample of 1,599 men, salivary testosterone showed no clear association with overall sexual function, erectile difficulties, lack of enjoyment or lack of sexual interest. And in a 2024 intensive within-person study, 41 men aged 18 to 26 gave daily saliva samples and daily desire reports across 31 consecutive days, 759 observations in total, and there was no positive relationship between a man's own testosterone and his own desire, either same-day or time-lagged. The small cross-effect among partnered men ran negative: higher testosterone predicted lower subsequent desire. Testosterone did positively predict courtship effort among single men on days they interacted with potential partners, and the authors flag that finding as exploratory.
Within the normal range, a higher testosterone number means more desire and drive.
A national probability sample of 1,599 men found no clear association between salivary testosterone and sexual function or sexual interest. A 31 day within-person study in 41 men, 759 observations, found no positive association between a man's own testosterone and his own desire, and a small negative cross-effect among partnered men. Both have real limits: one is cross-sectional and between-person, the other is undergraduate men in one industrialised country whose desire measure did not separate solitary from partnered desire. Neither shows the association the claim needs.
Threshold, or dose?
The tidy conclusion at this point is a pure threshold model: below some level things go wrong, above it more testosterone buys you nothing. That is tidier than the evidence. Bhasin's graded-dose work shows that lean body mass and muscle strength increase with testosterone dose, including across and above the physiological range. That is genuine dose-response, for one outcome.
Testosterone is one dial: push it higher and everything improves.
It depends entirely on which outcome. Dose-response holds for body composition, where graded and supraphysiological dosing raises lean mass and strength across and above the physiological range. Threshold behaviour is what the vitality, mood and desire data support, where correcting a real deficiency produced modest and partly missed effects and within-range variation predicts little. Different outcomes, different shapes. Collapsing them into one dial is the specific error that makes the retention argument look coherent.
Notice which side of that split the retention claims sit on. Body composition is the dose-shaped outcome, and it is not what anyone is retaining for. Mood and desire are the ones with threshold-shaped data. Focus, steadiness and confidence have no data here at all, in either shape. And none of the abstinence studies measured any subjective outcome whatsoever.
Two weak links in series is not a chain.
When to stop reading and see a doctor
One thing does not belong inside an argument about hormones. If low mood, pain, or erectile difficulty persists, that is a reason to see a doctor, and it is worth doing rather than putting off. It is not a reason to adjust a streak. Nothing in this article diagnoses anything, nothing in it predicts what will happen in your body, and none of it is a substitute for someone examining you.
So what is actually left
The argument requires two links. First, abstinence durably raises testosterone. Second, that raised testosterone is what makes you feel better. Link one rests on a withdrawn paper in 28 men and an uncounterbalanced study in 10, with nothing at all past three weeks. Link two, for the outcomes people care about, is unsupported within the normal range and mixed even in men who were genuinely deficient.
That is not the same as saying nothing happens when men retain. Plenty of men describe real changes and there is no good reason to call them liars. It says the hormonal explanation for those changes is not established, and that the obvious alternatives have never been separated out: expectancy, what a man does with the time and attention instead, sleep, and the plain effect of tracking anything at all daily. No study in this area has included a plausible active control.
The study nobody has run
Here is the gap, stated plainly: no study has ever measured hormones and daily subjective state at the same time, in men who are actually practising abstinence. Every paper above measured one or the other. That is the study the field needs and it does not exist.
There is a second gap that no study can ever close. Everything above is a between-person average, drawn from 10 men, or 41, or 790. An average tells you about a population. It does not describe you, and it does not forecast you.