Toxic grammar alert for Australians: Rodney Huddleston informs me that the ABC Radio breakfast show celebrated International Apostrophe Day on 16 August 2013 with disastrous results. Huddleston reports:
The presenter had brought in someone he called a grammar nerd/specialist and asked her about the use of the apostrophe. She managed to deal with dog's bowl and dogs' bowls, but when he asked her about children she said this was a collective noun, not a strictly plural and that in children's playgrounds and children's dreams the apostrophe should come AFTER the s.
I will not expose the grammar specialist's family to humiliation by naming her; I do have a heart. But this is really staggering misinformation. The apostrophe should never come after the s in cases of irregular pluralization. The genitive suffix is ’s unless the regular plural s immediately precedes it (in which case the genitive marker is simply the apostrophe alone). In irregular plurals like children, oxen, cacti, foci, phenomena, etc., there is no immediately preceding plural s, so the default holds: it's the children’s playgrounds, and likewise the cacti’s watering schedule, and these phenomena’s importance.
Beware of nonlinguists who appear on radio programs as grammar experts; they sometimes simply make stuff up.
The apostrophe is often spoken of as a punctuation symbol or grammatical marker, but I think it is best regarded as a 27th letter of the alphabet—a letter that has a required place in the spelling of various orthographic words but unfortunately has no specific phonetic correlate of its own, which makes it hard for people to remember where it goes (see this Huffington Post collection of photographs of spectacular spelling errors). Even for people who represent themselves to the media as grammar experts.
Double-checking the spelling uniformity against the familiar Wall Street Journal corpus of later 1980s newspaper text gave me a further shock, which certainly emphasizes the unpleasantly unmemorable character of spellings with apostrophes. Although the 732 occurrences of children's overwhelmingly confirm the rule, there are 13 distinct instances of childrens’ in the corpus!
One of them turns out to be simply commenting on the error:
And, inevitably, there are apostrophes appearing consistently at odd places: yours' or childrens' or did'nt. [File name: w7_038]
So that just confirms what I've said about the correct apostrophe placement: *childrens' is wrong. But the remaining dozen (plus repeats of three of them — the WSJ corpus unfortunately has some duplicated passages of text) are egg on face for the Journal: errors by WSJ writers that were (at least initially) missed by WSJ copy editors. This is perhaps sufficiently noteworthy that I should append the full text of the sentences below. I have done that, with the file name at the beginning of each.
w7_010: If parents come to associate national certification with the "best" instructors, they'll insist that their childrens' schools employ such teachers.
w7_064: The circus gala, a benefit for Emanuel Hospital's infants' and childrens' programs, was conveniently situated in a vacant lot next to the hospital.
w7_102: The government plans to establish a national curriculum and give parents greater say in their childrens' education.
w8_003: The transaction calls for the sale of Marvel's comic and childrens' book publishing, as well as licensing and merchandising operations.
w8_004: The company said yesterday that the preliminary agreement with Andrews Group calls for the sale of Marvel's comic and childrens' book publishing, as well as licensing and merchandising operations.
w8_031: Municipal bonds designed to help parents save for their childrens' college education are winning favor among state lawmakers, bond analysts and investors.
w8_039: The publishers have also sought a new "literary" look — more childrens' favorites and classics, less in-house writing and abridgment.
w8_078: The idea of early intervention also has become an issue in the 1988 presidential campaign, with candidates from both parties calling for more spending on childrens' programs. [Also in w8_116]
w8_083: Until recently, Troll Books was known primarily in the childrens' book business — and it still isn't well known to the general public. [Also in w8_120]
w8_085: The group, Friends for Education Inc., of Daniels, W.Va., said the four companies are providing data to parents that makes their childrens' performances on standardized tests appear better than they are relative to others' performances. [Also in w8_122]
w9_020: In addition to naming Mr. Keating, the 160-page suit alleges wrongdoing by his wife, several of their children and their childrens' spouses, many of whom work for American Continental or its Lincoln unit, based in Irvine, Calif.
w9_036: Before the 1986 tax act, many parents made the election for their childrens' bonds, so the interest would avoid tax or be taxed at a low rate, says Kevin Duvall of Ernst & Young, CPAs.
When I pointed this out to Rodney Huddleston, he checked the COCA corpus (450 million words of American English) and found 101 occurrences of *childrens’ there (as against 24,493 occurrences of children’s).
By the way, notice what you are seeing here: descriptive linguists like Huddleston and me, firmly committed to describing the language as it is, alleging clear errors in respectable published prose.
But do not adjust your set. There is no conflict here. The rules for where you put apostrophes are strict and well-known spelling conventions. There is virtually no latitude in them. (You can perhaps spell things like P’s and Q’s as Ps and Qs if you really want, and some people still write Clive James’ books rather than Clive James’s books, as if the final s on proper names like James were the plural s; but that's just about all the flexibility there is.) There is no developing dialect split here; we are looking at slips. Mistakes.
Descriptive linguists base their claims on evidence about how English grammar works, not on dogma; but it is our considered judgment that the examples above do not form part of the evidence. They are better regarded as evidence that (sometimes even in published material) people make spelling mistakes in English. And in the case of the word child, the evidence suggests that the genitive plural of child is probably spelled incorrectly something like 0.4 to 1.8 percent of the time on average (and the rate must be much higher in first drafts, emails, and so on). Spelling errors with the small and phonetically nondistinctive letter ’ are apparently harder to spot than misspellings like *recieve with the 26 other letters (though it should be noted that there are 4 distinct occurrences of *recieve in the WSJ corpus).
So not everything you read in the papers is true, and not everything you read in the papers is grammatical, and in fact not everything you read in the papers is even data bearing on what's grammatical.
[I have added some material and fixed a few errors of wording or editing since this first went up. Thanks to Joan Maling for pointing out the duplicate examples among WSJ examples.]