What was cited before

Before Veale et al. (2015), BJU International, pooled clinician measurements from up to 15,521 men, anyone wanting a clinically measured figure had to use one study at a time. Each had its own sample and sometimes its own protocol, which is why older figures still circulating disagree with each other and with Veale.

What a single study can and cannot give you

A well-run individual study can be excellent evidence on its own terms - clinically measured, with a stated method, from a defined sample. What it cannot give you is the stability that comes from pooling many such studies together, because any one sample of a few hundred men carries sampling variation that a pooled estimate averages out. A small study can report a mean a centimetre or more from where a much larger pooled estimate later lands, purely from sample size, without either study being flawed.

Before 2015, a reader had no easy way to tell whether a given figure they encountered was typical of the small-study literature or an outlier within it, because there was no pooled estimate yet to check it against.

Wessells 1996 as one example of the type

A 1996 study by Wessells, Lue and McAninch measured men flaccid, stretched and erect in one session and is a good illustration of what pre-2015 evidence looked like: a real clinical measurement, a real and useful finding about the relationship between states, and a sample small enough that it was never trying to be the last word on a population mean. Its abstract reports 80 physically normal men with mean flaccid, stretched and erect lengths of 8.8, 12.4 and 12.9 cm, and concludes that "stretched length most closely correlated with erect length". That study still gets cited today, correctly, for what it actually established - the stretched-to-erect relationship - and not for a population average it was never designed to produce.

Why they varied from each other

Single studies conducted in different countries, different clinical settings, and different decades used measurement protocols that were broadly similar but not identical in every detail - bone-pressed or not, which line was measured, how erection was obtained. A method difference this small can shift an entire study's mean by an amount that swamps its own confidence interval, which is exactly the kind of discrepancy that made single pre-2015 studies hard to compare directly against each other.

Sample sizes in this earlier literature were also typically far smaller than the pooled total Veale later assembled, which widens the uncertainty around any one study's reported mean even before method differences are considered.

Why the pooled estimate replaced them rather than adding to the pile

A meta-analysis does not invent new measurements. It pools existing studies under a stated inclusion rule and reports the combined result, which is valuable precisely because it does the averaging-out that no single earlier study could do on its own, and because its inclusion rule - measurement by a health professional - filters out weaker evidence rather than treating every prior figure as equally trustworthy.

That is why Veale 2015 functions as the current default reference rather than one more entry on a list: it is built from the earlier single-study literature, not a replacement written from scratch, and reading about the earlier studies is what explains why some older figures still in circulation sit slightly apart from the pooled mean without either being wrong.

Why it is still worth knowing the earlier layer

A reader who only ever sees the pooled figure loses the ability to place an older citation when they run into one, and older citations still turn up constantly in articles, forum threads and casual references that predate the meta-analysis or simply never updated to it. Knowing that the earlier literature existed, what kind of evidence it was, and why it varied more than the pooled estimate does is what lets a reader treat an older figure as a legitimate but less precise data point, rather than as either a fraud or an equal substitute for the current reference.

What this is not about

None of this touches how a figure should be judged once you have it - that is a separate kind of question, answered by a rating service rather than by a study, or by a human reviewer's impression rather than a clinical mean, or by browsing a public distribution of scores rather than reading a research pool. Nor does any of it involve estimating a length from a photograph, which an image model cannot do with the precision a study requires in the first place.