March Lith

December 30, 2015 7:35 AM

Randall Munroe observed a while ago: "In months other than September, the 11th is mentioned substantially less often than any other date. It’s been that way since long before 9/11 and I have no idea why.” David Hagen investigates. (The real culprit? A trend in font design from the late 1800s.)
posted by nebulawindphone (18 comments total)

this is vaguely like the xerox issue. smart approximations that turn out not to be so smart. i think we're going to see more of them as "artificial intelligence" takes off (maybe there's a trade-off - the cost of being smarter is making more mistakes). they're a bit like human mistakes, but because of the automation angle they become much more important (ie compare if one human read all the books in the world and then typed them into google's database with a consistently weird error with the more natural idea that many humans did the same, and so each error was less common). so perhaps the solution is to train multiple nets and divide work between them.
posted by andrewcooke at 7:51 AM on December 30, 2015 [1 favorite]

Even given the typography stuff discussed, I find it kind of hard to believe that there was only a tiny uptick in usage around the original Armistice Day/Remembrance Day and little mention since. I mean, the 11th hour of the 11th day of the 11th month is still pretty well observed here in Canada and I would guess, in Britain. Maybe I'm not understanding something.
posted by bonobothegreat at 7:59 AM on December 30, 2015 [1 favorite]

The largest open question is why nth was chosen so often. It seems like such a strange error to make. The word nth is a legal word in mathematical and scientific publications, so that should help its chances of getting picked. In most fonts the top of the n is really thin, and is likely invisible in many texts on which they trained the algorithm. But there is a big different in height between 1 and n, especially in the typewriter era, which is where the errors occur.

There isn't a very large difference in height between n and 1 if you use text figures, which you often see in older books.
posted by Monday, stony Monday at 8:05 AM on December 30, 2015 [7 favorites]

(text figures are also called lowercase numbers)
posted by Monday, stony Monday at 8:28 AM on December 30, 2015 [3 favorites]

January nth is coming! We should have a party.
posted by miyabo at 8:34 AM on December 30, 2015 [1 favorite]

I find it kind of hard to believe that there was only a tiny uptick in usage around the original Armistice Day/Remembrance Day and little mention since.

I think the issue is that most references are to "Armistice Day" or "Remembrance Day" or "Veterans Day". On the other hand, "September 11th" is still one of the major ways to refer to it (behind "9/11").
posted by Etrigan at 8:36 AM on December 30, 2015

Is there an explanation for the axes in the first figure, which plots count against frequency? I think count is the median numbers of hits, but I thought frequency was count, so it is a bit confusing. Anyone know?
posted by a lungful of dragon at 8:51 AM on December 30, 2015 [1 favorite]

I'm pretty sure the y axis is "count of days at a given frequency". So the 11th is on its own at around 1.6 and there are seven days with a frequency around 2.3 (the highest bar). This took me a while to puzzle out, honestly.
posted by paper chromatographologist at 9:03 AM on December 30, 2015 [2 favorites]

Ah, thanks!
posted by a lungful of dragon at 9:08 AM on December 30, 2015

I'm kind of amazed 711 or 7/11 isn't in there.
posted by nevercalm at 9:09 AM on December 30, 2015

Humans are prone to the same kind of errors. In a previous life, I came up with part numbering schemes for different product lines in the automotive aftermarket. I was never able to convince marketing or product engineering that our part numbering schema should forbid use of 1, L,l, I,i, and 0,O,o. (and dashes and spaces)
posted by klarck at 9:29 AM on December 30, 2015 [5 favorites]

I hate to make the 12th comment -- there should only be 11 -- but this is one of the most interesting things I've read in a while. Thanks for the post.
posted by LeLiLo at 10:19 AM on December 30, 2015 [1 favorite]

Yeah, but... this month goes up to 11.
posted by Pallas Athena at 10:55 AM on December 30, 2015 [2 favorites]

December seems off in Munroe's calendar. 31's font size is larger than 25's. The 31st should be large, but the 25th should be absolutely dominant.

Similar for February. The 14th blends in with the other numbers when it seems like it should be bigger. Maybe his sample of books had very few holiday mentions.
posted by honestcoyote at 11:43 AM on December 30, 2015

I think it's because people just say Christmas rather than saying the date. (To the extent that my husband has to ask me every year what day Christmas is.)
posted by Daily Alice at 11:54 AM on December 30, 2015 [3 favorites]

AI eggcorns! Wait until they start auto-generating spam with it in and humans pick up nth as a slang term for 11th.
posted by lucidium at 2:56 PM on December 30, 2015 [1 favorite]

So interesting!
posted by JHarris at 3:01 PM on December 30, 2015

That was a fascinating little tidbit. I'm mildly surprised Randall Munroe didn't delve into it himself; it seems like the sort of thing he'd have been interested in and investigated.
posted by Scattercat at 11:26 PM on December 30, 2015