Methods

The scale

Every bar is drawn to one scale in both directions. Its length is the published length. Its thickness is the real diameter, worked out as circumference ÷ π. Nothing is exaggerated to make bars easier to see.

Confidence tiers

• A — peer-reviewed, measured by a clinician (or a published clinical definition), and the figure has been traced to the paper itself.
• B — peer-reviewed but self-reported; or derived from a study's released raw data; or a modelled estimate from a meta-analysis; or a pooled aggregate from calcSD; or a manufacturer's specification.
• C — press-reported, company- or retailer-sourced, approximate historical figures, or a citation that has not yet been checked against its source.

The tier describes the length figure. How the girth was obtained is shown separately.

Measurement conventions

Bone-pressed (BP) means the ruler is pushed through the fat pad until it meets the pubic bone. Non-bone-pressed (NBP) means it rests on the skin. Across every study calcSD indexes, the pooled gap is 1.87 cm erect, 1.69 cm stretched and 0.57 cm flaccid. Ten studies measured the same men both ways; their within-study erect gaps run from 1.10 cm to 2.85 cm and cluster near 2 cm.

Stretched length is the flaccid penis pulled to full extension. It tracks erect length on average but is a different measurement, and in Western samples it runs about a centimetre longer than erect.

Several of the most-cited sources — including Veale 2015's nomograms — pool both conventions in an unstated ratio. Those rows are labelled "mixed BP/NBP" and cannot be converted. Veale 2015 states that its studies measured with the fat pad pushed to the bone, but most of the studies it lists do not say how they measured, so its rows are labelled mixed rather than bone-pressed.

Girth: measured, derived, extrapolated

• Measured — the source published a circumference.
• Derived (mean + SD) — a percentile computed from a published mean and standard deviation.
• Band midpoint — the source gave a range; the midpoint is shown.
• Extrapolated — no girth was published. It is estimated by percentile: a length at the 80th percentile is given the girth at the 80th percentile, using Veale 2015's means and standard deviations (erect length 13.12 ± 1.66 cm, erect circumference 11.66 ± 1.10 cm; flaccid 9.16 ± 1.57 cm and 9.31 ± 0.91 cm).
• Extrapolated beyond the data — the same calculation more than three standard deviations from the mean, where it stops describing anything real. Drawn faintly.

Length and girth are only weakly correlated in real men, and girth varies less than length in absolute terms. Extrapolation is therefore roughly right in the middle of the range and increasingly fictional at the extremes.

The normalised view

Rows measured without the bone press are shifted to their bone-pressed equivalent by the pooled gaps above. Where a study published both figures for the same men, its own bone-pressed figure is used instead (Wessells 1996: 12.89 cm NBP, 15.74 cm BP). Rows with mixed conventions are shown in place and flagged; rows whose convention was never stated are left out.

The every-study view

Each line is one dataset from calcSD's public list that publishes both a mean and a standard deviation. The dot is the mean. The whisker spans two standard deviations either side — a normal curve fitted to that study's own numbers, not an observed range. Studies drawn from erectile-dysfunction patients, self-reported samples and pooled aggregates are flagged by colour.

What gets a row

This is not a meta-analysis. A measurement earns a bar in the main views if it anchors a category nothing else covers — a method, a convention, a population, a product class — or is the best-powered example of its category, or demonstrates a failure mode worth seeing (selection bias, self-report inflation, convention ambiguity). A study that only adds another estimate of something already anchored is left to the every-study view.

Units

Inches are the drawing unit. Centimetre columns in the data are converted from the inch figures and may differ from the published centimetre value by up to 0.02 cm; the published value appears in the source line where the source gave one.