Bone-pressed in the literature

Fewer than people assume: much of the published literature measures from the skin at the base, not from the pubic bone. Some studies pooled into the standard reference figures specify bone-pressed measurement, some specify skin-to-tip, and some do not say which - a real gap between the assumption and the record.

What "bone-pressed" changes, briefly

The pubic fat pad in front of the bone varies in thickness between men, and pressing the ruler firmly into the bone removes that tissue from the reading. Not pressing leaves it in. The two methods are measuring genuinely different things - not the same length read two ways, but two different starting landmarks that happen to sit close together on most men and further apart on others.

The mix inside the pooled data

Veale et al. (2015) pooled studies from multiple sites and periods, and methods sections across that period are not uniform on this point. Habous et al. (2018) put it bluntly: "Much of the published data report penile length measured from the penopubic skin junction-to-glans tip" rather than from bone. Some clinical studies explicitly describe pressing to the pubic bone. Others describe placing the ruler at the base without specifying pressure, which is closer to a non-bone-pressed protocol, and a smaller number leave the detail out of the published method entirely. The meta-analysis inherits whatever its component studies did - it cannot retroactively standardise a measurement that was already taken.

What direction this pushes the pooled mean

A study that did not press to the bone tends to report a shorter length than one that did, because fat-pad tissue left in place covers part of the shaft the bone-pressed protocol counts. The gap can be large: in Habous et al. (2015), 778 men averaged 12.53 cm skin-to-tip and 14.34 cm bone-to-tip, erect, measured in clinic. Mixing both kinds of study into one pooled average does not average out to a clean bone-pressed figure - it sits somewhere between the two, pulled toward whichever protocol contributed more of the underlying sample. The direction of that pull is not stated cleanly in most summaries of the paper, which is a reasonable thing to be cautious about rather than to assume away.

What this means for placing your own reading

If you measured carefully, bone-pressed, and you are comparing that figure to a pooled percentile chart, you are comparing a clean bone-pressed number to a reference that is not entirely clean on the same dimension. That does not make the comparison useless. It means the chart carries a small amount of built-in uncertainty on top of its stated confidence interval, in a direction that would, if anything, make a careful bone-pressed reading look marginally larger against the pool than it would against bone-pressed-only data. Method differences of this kind are one of several reasons published studies disagree with each other, and this is a case where the disagreement sits inside a single pooled figure rather than between two separate papers.

Why this rarely gets flagged when a figure is quoted

A headline figure lifted from Veale 2015 - "the average erect length is X" - almost never carries a note about the underlying method mix, because the meta-analysis itself is reporting a pooled estimate across studies that, on this specific point, were not uniform. That is not a criticism of the paper, which is explicit within its own limitations about heterogeneity across the studies it pooled. It is a caution about the secondhand quoting of a single number stripped of that context, which happens constantly and loses the nuance every time.

Reading a chart with this in mind

Treat a percentile from any pooled chart as approximate at its edges rather than exact, and weight it a little less heavily the closer your own figure sits to a percentile boundary that a millimetre either way would cross. The method used to take your own reading belongs written down next to the number, for exactly this reason - it is the only way to know, later, which side of this gap your figure sits on.

None of this is a reason to distrust the pooled figures generally; a meta-analysis of 15,521 men is still the best reference available, and it is worth reading what that sample size does and does not buy you. It is a reason to hold a percentile loosely rather than treat it as a millimetre-exact verdict, which is a different failure mode from the one a rating service produces - a photo-based score from Rate Cock carries its own uncertainty, of a completely different kind, an AI model's estimate from an image being a recognition problem rather than a measurement one, and a human judge's opinion being a subjective read rather than a landmark-based reading at all - both worth naming so the measurement uncertainty here is not confused with either of theirs. A numeric score from a photo has its own error bars, and none of the three substitute for knowing what protocol produced the ruler figure you are holding.