What normal range means

"Normal range" in a results table is not an opinion about what counts as acceptable. It is a specific statistical construction: roughly the middle 95 percent of a measured sample, built from the mean and the standard deviation, and it means exactly that and nothing more when a paper uses the term.

How the range is built

Under the normal-curve assumption that pooled size data approximately follow, about 95 percent of a sample falls within roughly two standard deviations of the mean. A paper reporting a "normal range" is usually reporting that interval directly: mean minus about two SD to mean plus about two SD, or equivalently the 2.5th to the 97.5th percentile of the sample. Veale et al. (2015), BJU International, built its size nomograms this way, from "a weighted mean and pooled standard deviation" and a simulation of 20,000 observations from the normal distribution. For erect length, its pooled mean of 13.12 cm and SD of 1.66 cm put the 2.5th to 97.5th percentile interval at roughly 9.9 to 16.4 cm by that arithmetic. Either way of stating it describes the same middle slice of the distribution - it is a description of spread, built the same way any percentile interval is built once a reference population has been named.

The two ways of stating it - two SD from the mean, or the 2.5th to 97.5th percentile - coincide only under the normal-curve assumption, and it is worth knowing that is an assumption rather than a guarantee. Where a distribution departs from a normal curve, usually out in the tails, the two constructions can drift apart slightly, and a careful paper will say which one it actually used rather than treating them as interchangeable by default. Measured size data sit close enough to a normal curve that the difference is usually small, but "close enough" is doing real work in that sentence, not standing in for "identical."

Why "normal" does not mean what it sounds like

In everyday usage, "normal" carries a judgement - normal as in fine, unremarkable, nothing to note. In a results table, "normal range" carries no such judgement. It is simply the interval that contains most of the sample, constructed by definition to exclude the highest and lowest 2.5 percent on each side. A figure sitting outside that interval is, statistically, unusual relative to the sample - it is not, by virtue of the statistical term alone, anything else. Papers use the term in this narrow, technical sense, and a reader importing the everyday sense of "normal" onto a statistical range is reading a different claim than the one being made.

This is a common enough mix-up that it is worth naming directly: the everyday sense of "normal" describes a judgement about a single case, made by a person, against some standard of what is fine. The statistical sense describes a property of a whole distribution, built by arithmetic from a sample, and it exists whether or not anyone judges any individual figure as fine or not. A table can report a normal range for a quantity without that quantity carrying any judgement at all - it is simply reporting where the middle of the data sits, the same way it would for a measurement of almost anything else that varies naturally between individuals.

Keeping the two senses apart

This distinction matters because the same two words describe a precise statistical object in a methods section and a vague reassurance in ordinary conversation, and the two uses get blended without anyone noticing the swap. A results table stating a normal range has told you about the shape of a distribution - mean and spread, expressed as an interval - and has not told you, and is not designed to tell you, anything beyond that. This post covers only what the term means statistically as it appears in a table; it is not the place to work through the wider arithmetic of how many standard deviations bound what share of a distribution, which is covered separately.

If you are placing your own reading against a table like this, the number that matters is where you fall inside the stated interval, and getting a reading you can trust in the first place starts with taking the measurement the way the studies do. None of this transfers to a subjective score - Rate Cock, a numeric rating tool, a human judge's opinion, or an AI model reading a photo do not report a normal range in this statistical sense, because none of them is drawn from a measured sample with a mean and a standard deviation behind it.