CI or SD

A published table often carries two different measures of spread next to the mean, and they answer different questions. Mixing them up is one of the more common misreadings of a paper like Veale et al. (2015).

Standard deviation: about individuals

The standard deviation (SD) describes how spread out the individual measurements in the sample are around the mean. A larger SD means men vary more from each other; a smaller SD means they cluster more tightly. If you want to know how unusual your own measurement is relative to other men, the SD is the figure you need.

Confidence interval: about the mean

The confidence interval (CI) describes how precisely the study has pinned down the average, not how spread out individuals are. A 95% CI of, say, a few tenths of a centimetre around a reported mean says the true population average almost certainly sits within that narrow band - it says nothing about how far any one man sits from that average.

Why the two get swapped

Both are ranges reported next to a mean, formatted similarly in a results table, which is exactly why readers confuse them. A narrow confidence interval - the kind a large study like Veale 2015 produces - gets misread as "everyone is close to the average," when it actually only says "the average itself is known precisely." The standard deviation is the number that speaks to how spread individuals are, and it is typically far wider than the confidence interval around the mean in a study this large.

The practical rule

If a table entry is meant to tell you where the population mean sits, it is a CI or a standard error. If it is meant to tell you how one measurement compares to the range of other people, it is the SD. Reading a narrow CI as evidence that individual variation is small is the specific error worth avoiding.

For the fuller version of this distinction, worked through with the Veale sample size directly, see what 15,521 men buys you, which covers why a large sample tightens the mean without narrowing the spread. For the standard deviation itself, what it represents and how to read one against a single figure, see standard deviation, explained with length. The same distinction has no bearing on a photograph-based judgement - a score from Rate Cock is not accompanied by either statistic, because it is not an estimate of a population parameter. An image model's output carries no confidence interval in this sense either, since it has no measured sample behind a single photograph to compute one from; a ranked score on Penis Rater and an opinion from a Rate Penis judge are both single assessments, not statistical estimates with their own uncertainty range.