The mean is not the whole test
On this page10
The three numbers are not a sequence and not a range
Nineteen microns is an average. Thirty microns is a boundary for a sub-population. Twenty four percent is a limit on dispersion. They answer three different questions about one batch of fiber.
The average alone would be easy to satisfy badly
A blend of very fine and very coarse fiber can average under nineteen microns while containing a substantial fraction that nobody would call cashmere. The tail condition exists to prevent that.
So the tail is written as a weight fraction of a diameter class
No more than 3 percent by weight of cashmere fibers with average diameters that exceed 30 microns. Note that even the tail members are described by their own average diameter rather than by a single measurement.
Which means the rule contemplates averaging twice
Once across the whole fiber content to get the mean, and once per fiber to place it inside or outside the thirty micron class. Two different averages, both called an average diameter.
And the variance condition is permissive in its wording
The average fiber diameter may be subject to a coefficient of variation around the mean that shall not exceed 24 percent. May be subject to, not shall not exceed, and then shall not exceed inside the same sentence.
That reading matters because it allows a spread rather than requiring one
The paragraph grants tolerance. It says a measured mean need not be exact to qualify, within a bounded amount of variation, which is a different posture from a hard ceiling.
The species condition is the one that cannot be measured at all
Fine dehaired undercoat fibers produced by a cashmere goat, capra hircus laniger. No micrometer answers that. It is a provenance condition sitting alongside three measurements.
And dehaired is doing work in that phrase
A cashmere goat produces both a coarse outer hair and a fine undercoat. The rule is not describing the animal’s fiber, it is describing the undercoat after the outer hair has been separated out.
Four conditions, four kinds of evidence
A species, a mean, a tail and a spread. Which is a far more demanding structure than any other defined term in this library, and all of it attaches to permission to print one word.