Sunday, May 27

I keep walking back and forth to the kitchen, looking for a substance to consume which will quiet my appetites sufficiently that I can sit still and write. So far I’ve gone through a fair amount of yogurt, several pieces of bread with olive oil, a cup of coffee, and half a bottle of Grolsch. There’s no whiskey or tobacco in the house, so I suppose this will have to do. If I can get on a roll before the beer is gone, it may be enough.

I spent part of today tidying up a little piece of code I wrote months ago, so that Levi can perhaps show its results to people at a conference in Belgium. What it does is to grep through a bunch of transcribed US Census records from Nebraska in 1900 (available from this friendly ftp site as fixed-width text files) looking for groups of people who might possibly be twins (or even, say, triplets). It’s not all that clever about how it does this, and the results are correspondingly modest, but I suppose it’s a good enough proof of concept. You can find this kind of thing if you go looking for it.

Would you like to see some twin pairs? I thought you might.

last first relationship* sex born father's birthplace mother's birthplace
Airhart Lewis H. son M Jun. 1899 Illinois Germany
Airhart Luella H. daughter F Jun. 1899 Illinois Germany
Jordan Wallace P. Son M June 1897 Indiana Germany
Jordan Wallis C. Son M June 1897 Indiana Germany
Kaufman Minnie daughter F Dec 1886 Pennsylvania Illinois
Kaufman Vinnie son M Dec 1886 Pennsylvania Illinois
Phelps Rella V daughter F Nov 1886 Ohio Iowa
Phelps Frilla J daughter F Nov 1886 Ohio Iowa

They really did name their kids like that.

Out of a hundred odd pairs in my results, I count 25 whose names are obvious transformations/variations (Leonora/Elenora, Edgar/Edwin, Leonard/Lester) or rhymes (Minnie/Vinnie, Lerrie/Wernie), and 15 more starting with the same letter or syllable (Frank/Fred, Howard/Hazel, Roger/Ronald). Admittedly this is an arbitrary distinction - what constitutes a transformation? - but the pattern does stand out.

If you looked at a representative sample of present-day twins in America, would you see the same pattern? Maybe, but maybe this one expresses some kind of fundamental attitude we’ve since lost. I remember hearing in a history-of-the-family course that cultures with high infant mortality rates (most of them, historically) often treat naming much differently. Very young children might not be named for their first year or two, or until it seems likely that they’ll survive, and children born after the death of a sibling often take the same name - acting as literal replacements for the missing child.

I can imagine that a culture which understood twins as sharing an identity would give them tightly-coupled names, to an extent that seems sort of comical now but once just expressed an apparent truth. ‘course, it’s probably dangerous to talk about “a culture” here - Nebraska in 1900 was full of Czechs, Germans, Scots, Irishmen, Swedes, Russians, and sundry migrants from back East.

Somebody somewhere has probably written a dissertation on this.

There’s interesting, albeit unsourced, stuff on naming conventions at genealogy.com: The Importance of Given Names.

I can’t distribute the big wad of census transcriptions I’m using for this, but attached is a little tarball with the script and some output. If you happen to download a metric crapload of the right text files and cat them all together into something like 1900.total.txt, you can then do

perl twins1900.pl < 1900.total.txt > 1900.log

and be on your merry way. I’d provide more detailed instructions, but I kind of doubt it’s worth anyone’s time at this point. We want to make a lot of this kind of data available from a solid database with a friendly interface and robust search tools, but there are probably going to be some non-trivial hoops to jump through on the way.