Welcome to Feiba Peveli: Procedurally Generated Place Names from the 1820s

While reading the lovely book Paradise Now: The Story of American Utopianism by Chris Jennings, I came across this intriguing paragraph in the chapter on the New Harmony movement of the early 19th century:

The farmers christened their breakaway community Feiba Peveli. This peculiar name derives from a “rational system of nomenclature” invented by Thomas Stedman Whitwell, the British architect… Under Whitwell’s scheme, lines of latitude and longitude are associated with certain letters of the alphabet, so that a place’s name encodes its geographic coordinates. The results tend toward the tongue-twisting. New York City becomes Otke Notive; London is Lafa Vovutu; and New Harmony is Ipba Veinul.

This struck me in particular because I had moments earlier been playing No Man’s Sky, which gives its stars and planets random plausible-sounding names based on their three-dimensional positions in a consistent generated galaxy:

Screenshots courtesy XSurfer on Steam and Video Game Source on Youtube.

I was curious to see if I could track down the details of Whitwell’s original naming algorithm. It appears it was originally published in the “New Harmony Gazette,” a newspaper of the utopian community which Feiba Peveli spun off from: I’ve found references to this paper existing in microfilm collections, but none near me nor online as far as I can tell.

I did, however, find this gloss of the system in a scan of the 1902 book “The New Harmony Communities” by George Browning Lockwood (available on Google Books):

The stated rationale here, deploring the burgeoning repetition in American place names, reminds me fondly of the solutions to non-problems so often seen in academic games research, like “if only writers had a way to encode their ideas as directed planning graphs.” (I kid, to any friends reading this who are working on that.) Unique, utilitarian place names are both delightfully utopian and completely unnecessary. But I love name generators, and it seems like there’s enough information here to recreate this one, along with a number of examples (crucially, with coordinates) to test it on.

It immediately becomes clear, however, that the gloss is not accurate. If all latitude numbers are vowels and all longitude numbers are consonants, then you’d get extremely lopsided words. Feiba Peveli, at 38.11 N 81.53 W, should be something like Ieiaa Pblf. Furthermore, the table indicates the letter e = 2, and yet while there are no 2’s in the coordinate there are 2 standalone e’s in the attested name. Clearly the algorithm for conversion has been imperfectly summarized.

Squinting a bit, though, it’s clear there’s some sort of relationship here: 3 = f, for instance. If we arrange the given key by the numbers in the coordinate, we can see the pattern:

So it looks like the labels on the original table ought to be “vowels” and “consonants,” and the algorithm is to alternate between them moving through the coordinate: following this we get Feiba Pali. We were told that a west longitude is indicated by a v, though not where to put it; but if we add it between the number and decimal place of the W coordinate, and give it an e so it’s not a standalone consonant, that gets us Feiba Paveli. We still have “pa” instead of “pe,” but perhaps the unpreserved “extensive rules for pronunciation and for overcoming various difficulties” explains this. Cool! Let’s try it on a different example.

Feiba P(e)veli = 38.11 N, 81.53 W.
Ipad Evenle = 38.12 N, 87.52 W.

Uh oh. How can 38.11 be Feiba and 38.12 be Ipad??

This time, we started on the vowel. Why? No idea. Nothing seems substantially different about the two coordinates to suggest this. Starting on the consonant would produce Feibe Pveele, which is certainly less elegant: perhaps the namer can choose which order to use? And actually, we’ve got two consonants back to back here (n + l), so maybe alternation isn’t required: perhaps the namer can simply move through the numbers picking a consonant or vowel at each step, as they prefer?

The position of the West-indicating v is also problematic: here, it seems to come between the two digits of 87 W, rather than more cleanly at the decimal boundary. Perhaps it, also, can appear anywhere within the word? Finally, we have another off letter: ei instead of e. I’m not sure where these discrepancies are coming from, but given the imperfect summary of the algorithm, perhaps the names have likewise been imperfectly preserved. (I can imagine some 19th century copyeditor valiantly trying to apply English spelling rules to these made-up words, unwittingly muddling the whole thing up: “e-i-v? No that simply won’t do…”)

Let’s try another: Ipba Veinul, 38.11 N, 87.55 W.

Aha! Looks like we’ve cracked it. We get more confirmation that the namer can freely alternate between consonants and vowels as they like, and that the v for West can go anywhere in the word. Also, it looks like the e after the v is only necessary if it’s not followed by a vowel, which makes sense.

Let’s try one more: Uhi Ovouti, 55.34 N, 4.3 W.

Well, crap. According to the algorithm we’ve reverse engineered so far, the name should be something like Ulik Ovi. Where is that h in Uhi coming from? There’s not even an h in the key! The second half is also completely baffling. We can make some progress if we realize that t and ou map to zero, and try adding padding digits:

We’re still a bit screwed here trying to get Uhi Ovouti. I would buy a rule like “h indicates a double letter,” so uu -> uh, but the 4 in 55.34 doesn’t seem to be represented at all: there’s no o or k. If we break the rule that the word boundary corresponds to the latitude/longitude division, the O in Ovouti could be the 4; putting the v at the start of the word would then get us to ou = 0. But the end is a mess: there’s no further o or k for the 4 in 4.3, and the ending ti would be 03, not 30. If the numbers were ordered 4003 you could get Ovouti, but that doesn’t seem a sensible way to indicate 4.3. If the rule is to zero-pad the middle to four digits, you’d eliminate the distinction between 4.3, 43.0, 0.43, etc. And this still doesn’t explain the missing letter in Uhi.

Working through the other given examples results in similar discrepancies, especially around vowels and missing numbers. Maybe the names were transcribed phonetically (or just sloppily) for the book? Maybe the missing “extensive rules for pronunciation and for overcoming various difficulties” involved changing or removing vowels in some situations to produce more satisfying syllables? That would seem to break the ability to convert a name back to coordinates, which Lockwood claims was one of the system’s explicit goals.

(This ability is a bit suspect to begin with, though: the diphthongs, for instance, would seem to make it impossible to distinguish whether Bei means 123 E (b, e, i) or 18 E (b, ei). On the other hand, while it’s not indicated whether the second part of each coordinate is minutes or hundredths of a degree, none of the examples have a value over 60, which suggests minutes: and the diphthongs are associated with 7, 8, 9, and 0. So there may be something more complex going on in the hidden extra rules that still allows for a two-way conversion.)

All these confusions aside, there are some charming things about the (possibly imperfect) version of the algorithm reconstructed here. The fact that the namer has freedom to decide which numbers should be consonants and which vowels restores some aesthetics and creativity to an otherwise drab and mechanical process. For instance, my hometown of Santa Cruz could be known (among other possibilities) as Fyree Bedtiv, Iyien Avedeouf, or, my favorite, Imree Vadeti. More interestingly, different people or groups could each have their own place name (all still indicating its geocoordinates, in theory) while all conforming to the same schema — try that with the metric system.

The algorithm can also be scaled up or down to talk about smaller or larger areas. At 40 degrees N or S, each minute of latitude is about a mile apart; we could talk about going over to Imree Vadeto (the west side of Santa Cruz), the Imie Vadet urban area or Im Vade county; or you could go over to a particular person’s house, Imreela Vadetido (and I hope they’ve inscribed that on the gate). It would be kind of cool to just write Imreela Vadetido on a postcard and have it get delivered to the right place. (OK, nine-digit zip codes basically do that already, but feel significantly less like casting a spell.) Heck, imagine a world where people’s proper names worked this way, so your name encoded the exact location of your birth!

Anyway, I like to imagine Whitwell would be tickled to know that two hundred years later someone is thinking about his system, and that something not too dissimilar is being used to name quadrillions of places in a rather utopian, if fictional, galaxy.

No Man’s Sky screenshot by the author.

Update: within a few hours of posting, Marc Moskowitz had tracked down the original article, which is available here — thanks, Marc! This makes a few things clearer. For instance, it instructs to avoid combinations of single vowels which would produce a diphthong, so the digits 23 could be df, di, or ef, but never ei: this helps with reversibility. Uhi Ovouti is revealed to be a misprint: the original article lists these same coordinates as producing Ulio Ovuoti. This looks a bit like the li -> h error you sometimes get with OCR, or tired typesetters. (However, the end would still seem to be fouled-up: the original gives the explicit example of 8° 7' becoming 0807; by that logic, 4° 3' should be 0403, O(v)uoit. Perhaps Whitwell, also tired, swapped 3 for 30 in his head when preparing this example.)