The numbers behind the sizes.
Every figure on this page carries its sample size, who did the measuring, and the paper it came from. Where the size scale and the data disagree, the page says so rather than picking one.
Read in
Every figure here is a measurement taken to somebody's convention. How this app measures, and where it differs
Why there is no single average here
Across 21 studies, the reported averages for flaccid length span 3.06–4.63 in. The honest statement is a range: 3.35–3.74 in.
Tabulated across twenty-one studies of flaccid size. Three of them are self-reported and the rest professionally measured; the two must not be read as one sample. The study averages disagree by more than twice the within-study standard deviation — so the interval, not the point, is the honest statement, and the interval is a field on this record rather than a figure in this sentence.
Across 21 studies, the reported averages for flaccid circumference span 3.06–4.63 in. The honest statement is a range: 3.54–3.94 in.
Circumference at the thickest point, from the same tabulation — which is not where this app puts the tape, so it is context and never a comparison. The same caution applies.
What the large reviews pool to
Two systematic reviews, read first-hand, each row carrying the number of men behind it rather than the number on the title page. Not one of them can place a reading: what each publishes is an error on its own pooled average, which says how sure the review is about that average and nothing about how far apart two men are.
Belladelli F, Del Giudice F, Glover F, et al. "Worldwide Temporal Trends in Penile Length: A Systematic Review and Meta-Analysis." The World Journal of Men's Health 2023;41(4):848-860.
doi:10.5534/wjmh.220203
Flaccid length3.43 in
3.21–3.63 in — confidence interval on the pooled mean, which is not a spread between men. 40,251 men of the review's 55,761, from 33 of its 75 studies. Nothing here can produce a percentile.
A meta-analysis pooling studies published across eight decades. Its interval is an interval on the pooled mean and not a spread between men, so nothing here can produce a percentile. Its subgroup tables are not imported, and the refusal is recorded under its own name.
Stretched length5.09 in
4.91–5.27 in — confidence interval on the pooled mean, which is not a spread between men. 44,300 men of the review's 55,761, from 64 of its 75 studies. Nothing here can produce a percentile.
The same pooled analysis, on the dimension it rests most heavily on. Its interval is an interval on the pooled mean. Its subgroup tables are not imported, and the refusal is recorded under its own name.
Erect length5.48 in
5.2–5.77 in — confidence interval on the pooled mean, which is not a spread between men. 18,481 men of the review's 55,761, from 20 of its 75 studies. Nothing here can produce a percentile.
The erect row rests on a third of this review's participants and on the fewest of its studies, which is the same disparity the study bank shows within one paper. Its interval is an interval on the pooled mean, and its landmark is nowhere stated — so it cannot be placed beside a reading taken at the skin without a claim nobody has made.
Mostafaei H, Mori K, Katayama S, et al. "A systematic review and meta-analysis of penis length and circumference according to WHO regions: who has the biggest one?" Urology Research and Practice 2025;50(5):291-301.
doi:10.5152/tud.2025.24038
Flaccid length3.63 in
0.09 in — standard error on the pooled mean, which is not a spread between men. 28,201 men of the review's 36,883, from a review of 33 studies which does not say how many stand behind this row. Nothing here can produce a percentile.
A random-effects meta-analysis. Its published spread is a standard error on the pooled mean, which shrinks as studies are added and is not a spread between men — read as a standard deviation it would move a mid-range reading by more than ten percentile points. The subgroup tables this paper is organised around are not imported, and it states no measurement landmark for the pooled figures.
Stretched length5.06 in
0.13 in — standard error on the pooled mean, which is not a spread between men. 20,814 men of the review's 36,883, from a review of 33 studies which does not say how many stand behind this row. Nothing here can produce a percentile.
The same random-effects pool. Its spread is a standard error on the pooled mean and cannot produce a percentile. The subgroup tables this paper is organised around are not imported.
Erect length5.45 in
0.37 in — standard error on the pooled mean, which is not a spread between men. 5,669 men of the review's 36,883, from a review of 33 studies which does not say how many stand behind this row. Nothing here can produce a percentile.
The erect row rests on a small fraction of this review's participants, and its standard error is by far the widest of its five — which is the disagreement between its component studies showing through rather than the variation between men. It states no measurement landmark.
Flaccid circumference3.58 in
0.05 in — standard error on the pooled mean, which is not a spread between men. 30,117 men of the review's 36,883, from a review of 33 studies which does not say how many stand behind this row. Nothing here can produce a percentile.
A pooled circumference. Its spread is a standard error on the pooled mean. Where on the shaft its component studies put the tape is not stated, and the field measures girth at the mid-shaft, at the base and at the thickest point under one word — so this is context and never a comparison.
Erect circumference4.69 in
0.07 in — standard error on the pooled mean, which is not a spread between men. 5,168 men of the review's 36,883, from a review of 33 studies which does not say how many stand behind this row. Nothing here can produce a percentile.
The same pool, on the dimension it rests least heavily on. Its spread is a standard error on the pooled mean, and its measurement site is not stated.
The five sizes, and what they really hold
The scale is designed so that half of men land in Medium and one in twenty at each end. The right-hand column is what the study actually puts in each band at those exact cut points.
Flaccid length
10,704 men, measured by a health professional.
| Size | Range (in) | Scale says | This study |
|---|---|---|---|
| Micro · XS | up to 2.2 | 5.0% | 1.0% |
| Small · S | 2.2 – 3.1 | 20% | 22% |
| Medium · M | 3.1 – 4.1 | 50% | 57% |
| Large · L | 4.1 – 5.1 | 20% | 19% |
| Hung · XL | 5.1 and over | 5.0% | 0.7% |
- Micro: The scale calls this 5% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is 1%.
- Hung: The scale calls this 5% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is 1%.
Erect length
692 men, measured by a health professional — a small sample for a percentile, and for the tails especially.
| Size | Range (in) | Scale says | This study |
|---|---|---|---|
| Micro · XS | up to 3.9 | 5.0% | 3.0% |
| Small · S | 3.9 – 5.1 | 20% | 44% |
| Medium · M | 5.1 – 6.3 | 50% | 49% |
| Large · L | 6.3 – 7.5 | 20% | 4.1% |
| Hung · XL | 7.5 and over | 5.0% | <0.5% |
- Micro: The scale calls this 5% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is 3%.
- Small: The scale calls this 20% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is 44%.
- Large: The scale calls this 20% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is 4%.
- Hung: The scale calls this 5% of men. In Veale D, Miles S, Bramley S, Muir G, Hodsoll J's sample it is under 0.5%.
Flaccid circumference
9,407 men, measured by a health professional.
| Size | Range (in) | Scale says | This study |
|---|---|---|---|
| Micro · XS | up to 3.1 | 5.0% | 5.0% |
| Small · S | 3.1 – 3.4 | 20% | 20% |
| Medium · M | 3.4 – 3.9 | 50% | 50% |
| Large · L | 3.9 – 4.2 | 20% | 20% |
| Hung · XL | 4.2 and over | 5.0% | 5.0% |
These cut points were computed from this study to hit the scale's intended shares, so the two right-hand columns agree by construction. That agreement is a tautology, not a finding — unlike the length bands, which are round guideline numbers and miss.
Erect circumference
381 men, measured by a health professional — a small sample for a percentile, and for the tails especially.
| Size | Range (in) | Scale says | This study |
|---|---|---|---|
| Micro · XS | up to 3.9 | 5.0% | 5.0% |
| Small · S | 3.9 – 4.3 | 20% | 20% |
| Medium · M | 4.3 – 4.9 | 50% | 50% |
| Large · L | 4.9 – 5.3 | 20% | 20% |
| Hung · XL | 5.3 and over | 5.0% | 5.0% |
These cut points were computed from this study to hit the scale's intended shares, so the two right-hand columns agree by construction. That agreement is a tautology, not a finding — unlike the length bands, which are round guideline numbers and miss.
The published distributions
Percentiles reconstructed from each study's own mean and standard deviation. This is exact rather than approximate for these nomograms: the review built its published curves by simulating 20,000 observations from a normal distribution. Where a second row appears, it is not the paper's — it is the same curve moved by the stated offset below, which is the distribution this app actually places a reading in.
Flaccid length
10,704 men, measured by a health professional.
| 5th | 25th | 50th | 75th | 95th |
|---|---|---|---|---|
| 2.59 | 3.19 | 3.61 | 4.02 | 4.62 |
Stretched length
14,160 men, measured by a health professional.
| 5th | 25th | 50th | 75th | 95th |
|---|---|---|---|---|
| 3.99 | 4.71 | 5.21 | 5.71 | 6.44 |
Erect length
692 men, measured by a health professional — a small sample for a percentile, and for the tails especially.
| 5th | 25th | 50th | 75th | 95th | |
|---|---|---|---|---|---|
| As published | 4.09 | 4.72 | 5.17 | 5.61 | 6.24 |
| As sizr compares it | 3.38 | 4.02 | 4.46 | 4.9 | 5.53 |
These figures are read as bone-pressed — the ruler pressed through the fat pad to the bone. A reading taken at the skin, which is what this app's own tape reports, is placed against the second row instead: the same curve with 0.71 in taken off the mean before the comparison. It is subtracted from this distribution only — circumference is compared as published.
That figure is derived, not read off a paper. No published series reports the gap as its own statistic, so it is the difference between two means printed for the same men, in two clinician series that measured both landmarks in one cohort:
- Habous 2015, 778 men — a gap of 0.71 in. Both landmarks taken in the same men, in one clinician series.
- Habous 2018, 201 men — a gap of 0.76 in. The same pair of landmarks in the same men, in a second series.
Salama 2018 reports a far larger gap and is not used. Its non-bone-pressed mean sits far below every other series in this bank, so what is unusual about it is where that series started the tape rather than how much the fat pad adds. A gap that large is not the same measurement done twice.
Neither series was read off its own paper: both figures were transcribed by somebody else, which is a class this page refuses to accept as a distribution and lists in the refusals below. That is why the offset is stated as this app's own assumption and carries no citation of its own — it adjusts a comparison, it does not join the data.
Flaccid circumference
9,407 men, measured by a health professional.
| 5th | 25th | 50th | 75th | 95th |
|---|---|---|---|---|
| 3.08 | 3.43 | 3.67 | 3.9 | 4.25 |
Erect circumference
381 men, measured by a health professional — a small sample for a percentile, and for the tails especially.
| 5th | 25th | 50th | 75th | 95th |
|---|---|---|---|---|
| 3.88 | 4.3 | 4.59 | 4.88 | 5.3 |
What one panel said it would choose
A stated preference is not a measurement of anybody. It is what a group of women picked out of a set of printed models, in one city, on one afternoon — and it is here because the question gets asked and an unsourced answer is worse than a sourced small one.
| Asked about | Length (in) | Circumference (in) |
|---|---|---|
| For one occasion | 6.42 | 5 |
| For a long-term partner | 6.3 | 4.8 |
60 women of the study's 75, rating both measurements in both situations — one panel, four averages, never four samples. They chose among physical models built in steps of 0.5 in, so nothing finer than one step was expressed by anybody here.
Women recruited in one university city, choosing among printed three-dimensional models rather than reporting on any partner. The models were checked against a tape after printing.
A stated preference is not a measurement of anybody, and this one rests on a subset of the women in the study rather than on its headline. The choices sat on a fixed grid of physical objects, so a difference finer than one step of that grid is not something anyone here expressed, and the paper prints no spread of any kind beside these means. A model has no pubic bone and no fat pad, so its length is the object's own — neither this app's chord nor a clinical landmark.
Prause N, Park J, Leung S, Miller G. "Women's Preferences for Penis Size: A New Research Method Using Selection among 3D Models." PLoS ONE 2015;10(9):e0133079. doi:10.1371/journal.pone.0133079
A clinical volume, which is a different quantity
Measured by ultrasound, one side at a time. The formula is on every row because the same testis yields three different volumes depending only on which constant the paper multiplied by — a volume without its formula is a number rather than a volume.
Both sides1.05 in³ ± 0.25
248 men, ellipsoid, length by width by height times 0.52.
Right1.092 in³ ± 0.269
248 men, ellipsoid, length by width by height times 0.52. Below 0.732 in³ the paper calls the testis small.
Left1.007 in³ ± 0.25
248 men, ellipsoid, length by width by height times 0.52. Below 0.671 in³ the paper calls the testis small.
A per-testis clinical volume, measured by ultrasound on one side at a time. This app measures the volume of the SHAFT and nothing inside the sac, so the two are different quantities that happen to share a unit and they never share an axis. Reading them as one would set a whole shaft against a single testis.
Healthy, fertile men in a multicentre ultrasound study run to one standard operating procedure, which states the volume formula it used.
Lotti F, Frizza F, Balercia G, et al. "The European Academy of Andrology (EAA) ultrasound study on healthy, fertile men: An overview on male genital tract ultrasound reference ranges." Andrology 2022;10(Suppl. 2):118-132. doi:10.1111/andr.13260
What would change these numbers
These are not disclaimers. Each one is a specific, checked fact that changes how a figure should be read, and the first is large enough to swamp several sliders in the builder.
These erect figures are read as bone-pressed, and a reading taken at the skin is not — so a stated offset comes off the mean before the two are compared.
A bone-pressed measurement presses the ruler through the pubic fat pad to the bone; a non-bone-pressed one starts at the body surface. The published gap between them is wider than the effect of most sliders in this app. Veale 2015 is closed-access: the review's own inclusion criterion describes a bone-pressed procedure, while the candidate studies it pooled are mixed, so the convention is asserted by the review rather than stated in every paper behind it. It bears directly on this app's own numbers, because sizr measures a CHORD from the proxy's base-ring centroid, which sits at the body surface: that is a skin measurement, and placed against a bone-pressed mean as it stands every percentile would read systematically LOW. So a stated offset is subtracted from the published erect-length mean before a skin reading is placed in it, and the figure is printed beside every comparison it changes. Circumference is not bone-pressed and is compared as published; the published row of every table on this page is the distribution exactly as the papers gave it, and the size bands are computed from the published mean unchanged.
- Commonly reported gap, low end: 0.39 in
- Commonly reported gap, high end: 0.79 in
The offset is 0.71 in, taken off erect length only. The size bands above, and the ladder behind the rest of the app, are computed from the published mean unchanged.
The percentiles assume a normal distribution, and two of the largest primary samples reject it.
Veale 2015 built its published nomograms by simulating 20,000 observations from the normal distribution, so reproducing them from a mean and an SD is exact rather than an approximation — that part is sound. But Ponchietti (n=3,300) and Nguyen Hoai (n=14,597) both reject normality by Kolmogorov-Smirnov on their own data. A pooled SD that also carries between-study heterogeneity makes the curve too flat, which pushes percentiles toward the middle. Treat a reading near the tails as softer than a reading near the median.
Self-reported figures run high — by more than the whole spread between men that this bank's erect distribution reports.
King 2021 measured the head-to-head gap between self-reported and researcher-measured erect length and put it above a standard deviation of the distribution this bank quotes; the figure itself is a field on this record rather than a sentence, so it prints in whichever units you are reading. That is why this bank carries no self-reported distribution and no pooled row across methods: averaging the two launders the bias into a figure that reads as authoritative.
- Self-reported minus researcher-measured erect length, King 2021: 0.83 in
Sources
Veale D, Miles S, Bramley S, Muir G, Hodsoll J. "Am I normal? A systematic review and construction of nomograms for flaccid and erect penis length and circumference in up to 15,521 men." BJU International 2015;115(6):978-986.
Flaccid length — n = 10,704, mean 3.61 in, SD 0.62 in. Pooled across 17 studies; measured by a health professional to a standard procedure. Excludes self-report, and excludes men with penile abnormality, previous surgery, erectile dysfunction, or a complaint of small size.
Veale D, Miles S, Bramley S, Muir G, Hodsoll J. "Am I normal? A systematic review and construction of nomograms for flaccid and erect penis length and circumference in up to 15,521 men." BJU International 2015;115(6):978-986.
Stretched length — n = 14,160, mean 5.21 in, SD 0.74 in. Pooled across 17 studies; measured by a health professional to a standard procedure. Excludes self-report, and excludes men with penile abnormality, previous surgery, erectile dysfunction, or a complaint of small size.
Veale D, Miles S, Bramley S, Muir G, Hodsoll J. "Am I normal? A systematic review and construction of nomograms for flaccid and erect penis length and circumference in up to 15,521 men." BJU International 2015;115(6):978-986.
Erect length — n = 692, mean 5.17 in, SD 0.65 in. Pooled across 17 studies; measured by a health professional to a standard procedure. Excludes self-report, and excludes men with penile abnormality, previous surgery, erectile dysfunction, or a complaint of small size.
Veale D, Miles S, Bramley S, Muir G, Hodsoll J. "Am I normal? A systematic review and construction of nomograms for flaccid and erect penis length and circumference in up to 15,521 men." BJU International 2015;115(6):978-986.
Flaccid circumference — n = 9,407, mean 3.67 in, SD 0.35 in. Pooled across 17 studies; measured by a health professional to a standard procedure. Excludes self-report, and excludes men with penile abnormality, previous surgery, erectile dysfunction, or a complaint of small size.
Veale D, Miles S, Bramley S, Muir G, Hodsoll J. "Am I normal? A systematic review and construction of nomograms for flaccid and erect penis length and circumference in up to 15,521 men." BJU International 2015;115(6):978-986.
Erect circumference — n = 381, mean 4.59 in, SD 0.43 in. Pooled across 17 studies; measured by a health professional to a standard procedure. Excludes self-report, and excludes men with penile abnormality, previous surgery, erectile dysfunction, or a complaint of small size.
What is not here, and why
Everything below was considered for this page and left off it. Several would be the most useful figures it could hold, which is exactly why they are named rather than quietly missing: the next person to go looking will find them too, and should find the reason with them.
Pooled cross-study aggregates published by size calculators
A pooled figure is not a study: it has no method, no population and no author to be wrong. The spread published beside these is built as a weighted average of the component studies' own spreads rather than as a pooled variance, which throws away the disagreement between them and makes every tail too extreme. Two of the sites the owner named publish the same aggregate, one of them by attribution to the other, so the field is smaller than it looks.
Mostafaei 2025, regional subgroup tables
The paper enters this bank as a pooled summary and its subgroup tables do not. This app publishes no ranking keyed by where somebody is from, and a review organised around such a table is the likeliest way one would arrive by accident.
Belladelli 2023, regional and temporal subgroup tables
The same rule, and the same reason for stating it beside the record rather than only in a decision: the pooled rows are imported and the tables the paper is organised around are not.
Country and continent averages
Not published here at all. Where the field prints these, the cells behind a label naming a dozen countries are often one small study, and self-reported and clinician-measured samples are mixed without saying which is which. Three of the seven sites read for this wave decline to publish such a ranking, two of them with their reasons stated.
Grower and shower band thresholds
Every threshold offered in the field is printed without a paper behind it, and one site's headline growth statistic divides two unrelated subsamples of a single review, which is not an individual's growth and cannot be. The RATIO is a real quantity this app can compute from its own two readings; the bands are not, so the ratio is what may be shown and the labels are not.
Condom stretch ranges and standard nominal widths
The stretch range and the standard nominal width are stated without a source everywhere they appear, and the largest product bank in the field self-rates most of its own rows at its lowest confidence with several carrying no link at all. A nominal width derived from a measured circumference, with the derivation printed, is arithmetic this app can do and stand behind; a range recalled from nowhere is not.
Size classification cut points published as fractions of a spread
Printed without a paper behind them, and in one case applied to circumference, where the clinical term borrowed for the smallest band has no meaning at all. This app derives its own cut points from a study's own distribution and says so, and prints the difference between what a scale intends and what a sample actually puts in each band.
Self-reported erect distributions
Self-measured, and in the largest of them measured along the underside, which is a fourth landmark and the reason a self-reported average lands near a clinician bone-pressed one. This bank carries no self-reported distribution and no pooled row across methods. THAT INCLUDES THIS SITE'S OWN MEMBERS. A figure somebody states in the builder is an ASK: it is solved into a body, the body is taped, and the ask is kept beside the tape so a member can see where the two disagree. It is never a sample, never pooled into a distribution, and reaches no percentile, no band share and no cut point published here — which is the same refusal as the rows above, applied to the one population this app could most easily collect and would be least entitled to publish.
Community surveys of self-reported size
A survey of a community that selected itself for the thing being measured. Its average sits far above every clinician series in this bank, which is what self-selection does and not what a population does.
Circumference measured at the thickest point
A maximum imported as a midpoint is a wrong number wearing a citation. This app measures a circumference at the mid-shaft, and the difference between the two sites cannot be corrected for across studies because the disagreement between studies is wider than the difference between the sites.
Circumference measured at the base of the shaft
A base circumference is systematically larger than a mid-shaft one, and several of the largest clinician series in the field measure there. They are refused as comparisons rather than as papers: a base measurement recorded as such, beside a mid-shaft one labelled as such, would be honest — pooling the two is not.
Lengths taken along the curve, or averaged between the outer and inner curve
This app reports a CHORD — the straight line from the base ring to the farthest point of the shaft — because a ruler from the pubic bone is what the dimension means clinically. An arc taken along the top follows the curve and is invariant under a bend; a mid-line averaged between the outer and inner curve is approximately the same thing. Either one imported as a chord would make the reported figure ignore the curvature it was measured under, and one site in the field argues explicitly against the straight rule this app uses. Where a definition is in dispute this app's own wins and the difference is stated rather than reconciled.
Figures filed under a bone-pressed key whose own method describes no pressing
An unstated landmark filed as a stated one is exactly what the provenance field on a record exists to catch, and it is worse than an unstated one: it produces a confident comparison out of a guess.
Samples including participants under age
A large flaccid sample in the field includes a substantial minority of participants below the age of majority. This bank is adult-only by design, as is the app around it.
Studies whose spread was never published, stored elsewhere as zero
The paper published no spread and the bank holding it records a zero, which turns every percentile against that record into arithmetic with exactly one answer. A missing spread is a kind of dispersion in this bank and it refuses a percentile rather than inventing one.
Papers whose results and discussion disagree about their own figures
Flagged by the aggregate bank that holds them, and confirmed as a reason not to import rather than a reason to pick one of the two numbers.
Preprints
One large recent candidate is a preprint, and separately measured its circumference at the thickest point rather than the mid-shaft. Either would be enough on its own.
Figures whose only surviving record is a marketing or archived page
A number whose only surviving source is a product page in a web archive has no method and no author. It is a lead, not a datum.
Ruler landmarks, everyday object dimensions and named individuals
Not one of the object dimensions published across four of the sites read for this wave arrives with a source, one site labels one of its own landmarks as invented, and two publish measurements attributed to named living people with no citation, method or provenance of any kind. An object may join this app's comparison only with its real dimension and where that dimension came from, exactly as a study does.
A preference average pooled across unnamed studies
One site publishes a headline preference figure pooled across studies it does not name, beside a correct citation for a different one. The named paper is imported and read first-hand; the pool is not.
Testicular volume means published without the formula behind them
Two rows labelled as different methods carry an identical average and an identical spread, which the formulas behind those labels cannot both produce, and none of the means on that page carries a source. A testicular volume without its formula is a number rather than a volume.
Clinician series recovered from an aggregator rather than read off their papers (Habous 2015 and 2018, Salama 2015, Park 1998, Wessells 1996, Khan 2012, Wu 1990, Nguyen 2021, Ponchietti 2001, Soylemez 2011, Shalaby 2014, Hwang 2005)
These are CANDIDATES, not rejects, and several would be the most valuable records this bank could hold — a number of them publish a length measured at the body surface, which is directly comparable to this app's own chord with nothing subtracted and nothing derived, and several publish both landmarks in the same men, which is the evidence for the offset itself. Every figure available for them here was transcribed by an aggregator rather than read off the paper, and that bank demonstrably carries at least one transcription error in exactly this class of record. Under this bank's standing rule a figure that has not been read off its paper does not get to look like one. Four publishers refused the reads attempted for this wave, so the reads have to happen somewhere else before any of these becomes a record. TWO OF THE NAMED SERIES ARE NEVERTHELESS SPENT, and pretending otherwise would be the omission this whole collection exists against: the pair that measured both landmarks is where this app's stated landmark offset comes from. What that refusal means and does not mean is settled on the offset record itself, which points back at this one by name — none of it enters as a distribution, none of it produces a percentile, and it carries no citation of its own precisely because this app has not read it.