Data
Heaping
Self-reported figures cluster at whole and half units far more than measured ones do, and the pattern itself is evidence of rounding, mostly upward.
Guides on Data: How a size study is built, Reading a nomogram without fooling yourself, The Veale meta-analysis, read properly
Self-reports pile up at whole numbers because people round when they type a figure, and they round toward the flattering side more often than the nearest one. The spikes are evidence of guessing, not of measuring.
Digit heaping is a known phenomenon in data collection, not something specific to this subject. Ask people to report their weight, their age at an event, or their income, and the answers pile up at round numbers - tens, fives, wholes - far more than a measured version of the same quantity would. Trained staff do it too: in a 2003 chart review in the Journal of Human Hypertension, Thavarajah and colleagues found zero was the final digit of 40% of systolic blood pressure readings taken by nursing staff. Self-reported length figures show the same pattern, and the shape of that pattern is itself a form of evidence about how the numbers were produced.
What "heaping" means on a histogram
A clinician-measured dataset, taken with a ruler that reads to the nearest millimetre, produces a histogram that is smooth: every value near the mean is about as likely as its neighbours, give or take sampling noise. A self-reported dataset, gathered by asking people to type or say a number, produces a histogram with spikes. Values ending in whole units and, to a lesser extent, half units, are reported far more often than a smooth curve would predict, while the values in between are reported less often than expected. The self-reported curve has not actually skipped those in-between values in the population - it has skipped them in the reporting.
Sketched schematically, the difference looks like this:
The tall bars sit at whole-unit marks; the short bars sit between them, where a real measured population still has plenty of people, but where a self-reporter is unlikely to type a figure.
Why the heaping runs one direction
Rounding, on its own, would heap symmetrically - a true value just above a whole number rounds down as often as a true value just below it rounds up, and the net effect on the mean should be close to zero. That is not what the pattern shows here. Reports cluster at round numbers, but the excess is concentrated on the high side of the plausible range for a given person rather than spread evenly, which is consistent with people rounding toward the more flattering neighbouring whole number rather than to the nearest one. King's 2021 review in the Journal of Sex & Marital Therapy puts a size on that pressure: often-cited self-report studies had means of about 6.2 inches for heterosexual men, while 10 studies where researchers did the measuring combined to 5.36 inches (n = 1,629). This is the same directional pressure that produces the broader self-report gap against clinician-measured figures, showing up here as a specific, visible fingerprint in the shape of the distribution rather than only in its mean.
Heaping is also a tool researchers use as a data-quality check independent of any external comparison. A survey that shows heavy heaping at round numbers is signalling low measurement precision even before anyone compares it to a separate clinician-measured dataset - the heaping is visible in the self-reported data alone, which is part of why it is treated seriously in survey methodology generally, not just here.
What this post is not
This is a description of a population-level pattern, not a checklist for your own habits when you write down a reading - that is a narrower, practical question with its own answer elsewhere. It is also not the same claim as "self-reported numbers run high," which is about the mean; heaping is a separate, complementary observation about the shape of the distribution, and a dataset could in principle show heaping without a shifted mean, or a shifted mean without visible heaping. Reading both together, alongside what a proper meta-analysis of clinician-measured data actually pooled, is how you tell a genuine population figure from a stack of typed-in guesses.
None of this describes a rating, a score, or a photograph-derived estimate - a heaped histogram is a property of numbers people typed into a form, which is a different object from a judged score or an image-based estimate, and it is a different object again from a distribution of ratings rather than lengths. A service like Rate Cock is answering that different, subjective question entirely, and its output was never a length in the first place, so it cannot heap or fail to heap in the sense this post describes.