The online survey problem

Anonymous online surveys produce figures above every clinician-measured study because they combine self-selection, self-measurement by an unspecified method, and no verification. That makes a survey a different instrument, not a cheaper clinical study, and comparing its output to Veale et al. (2015) as if the two were interchangeable is how bad numbers spread.

Who answers

A clinical study recruits a defined population - often men attending a clinic for an unrelated reason - and measures nearly everyone who is asked. An anonymous online survey recruits whoever finds the link and chooses to click through. That is a self-selecting sample, and the direction of the selection is not neutral: a survey about size is more likely to be completed by someone motivated to report a favourable figure, or simply by someone who thinks about the subject enough to seek the survey out, than by a representative cross-section of men. There is no comparable selection pressure in a study where a nurse measures whoever is next in the clinic.

How they measure

Nobody checks the ruler. There is no observer confirming bone-pressed technique, no standardised state, no landmark check, and often not even a request to specify the method used. The method itself decides more of the number than most people expect, and an online survey has no way to enforce or even record which method its respondents used.

What cannot be checked

A clinical study can be audited: the protocol is published, the measurer is named or described, the instrument is specified. An anonymous survey response cannot be checked against anything. There is no way to confirm the respondent measured at all, measured the state they claimed, or reported the number they actually got rather than a rounded or adjusted one. This is not an accusation of widespread dishonesty so much as a structural fact: a format with no verification step will always include some fraction of unverifiable answers, and there is no way to subtract that fraction back out afterward.

Why the result sits high

Put self-selection, unverified method and no check together and the expected direction is unambiguous: the reported figures land above what clinician-measured studies report, and by more than any single mechanism alone would produce. The gap is measurable: King et al. (2019) found a mean self-reported erect length of 6.62 inches among 130 sexually experienced college men, against a combined mean of 5.36 inches (13.61 cm, n = 1,629) across ten studies where researchers took the measurements, as King (2021) later pooled them. In the 2019 sample, higher social-desirability scores went with larger reported figures. This post is deliberately not re-deriving the specific mechanisms of self-report inflation - that list already exists and is worth reading in full - and it is not the same design question as a survey that offers a product sized to the reported figure, which adds an incentive on top of everything described here.

Comment sections and forum threads are the same design

Everything above applies equally to a number volunteered in a comment thread or a forum post, not just to a formal survey with a submit button. The mechanism is identical: whoever chooses to post a figure is self-selected, the figure is self-measured by an unspecified method, and nothing about the platform can verify it. A thread with a hundred posted figures is not a sample of a hundred men - it is a sample of a hundred people willing to type a number into a public forum, which is a narrower and differently biased group again. Treating a forum average as if it were survey data, and survey data as if it were clinical data, compounds the same problem at every step rather than adding a new one each time. The mismatch is often institutionalised, too - how a forum's own posting rules compare to an actual clinical protocol shows how far a community's conventions can drift from anything a study would accept.

What an online survey figure is actually useful for

It is a measurement of what a self-selected group of people typed into a form. That can be an interesting thing to know in its own right, but it is not comparable to the pooled clinician-measured figures from Veale 2015, and treating it as a rival dataset to that meta-analysis is a category error, not a difference of opinion about which source to trust.

None of this is an argument against subjective judgement as a category - it just is not this category. A rated photograph on Rate Cock is explicitly a judgement, not a self-reported measurement pretending to be one, and a scoring platform is upfront about that distinction. A human reviewer's opinion is the same kind of honest subjectivity, stated as an opinion rather than dressed as data, and an AI estimate from a photo carries its own accuracy limits that are worth reading on their own terms rather than assumed, described here.