Glyph of the word 'lavaka'.


  • (v.) to right someone’s path, to put someone on the right path
  • (n.) fixing something that’s crooked
  • (adj.) right, correct

A male lavaka ei ie ipe leleya kamelaye.
“I shall set that turtle on the right path.”

Notes: I was imagining a very little turtle that somehow started wandering towards the street… I’d fix that.

This post is coming pretty late. I’m totally backlogged (in general). Playing a little catch-up.

So regarding phonology, often one gets a sense that a certain phonological form sounds “right” for a given concept. There’s nothing wrong with that, per se. However, if one compares certain concepts in different conlangs, they often sound rather similar. This is because a lot of the conlangs that one encounters on the internet (or at least around these parts) are created by English speakers, and English speakers have inherent biases about how things “should” sound.

These biases, of course, are tendencies, not rules, and they’re a bit hard to quantify. However, if you ask 100 English speakers which name is best suited to a beautiful woman—Elara or Blurkf—all 100 (unless they’re being ironic or contrary) are going to choose Elara.

It’s to avoid these biases that a lot of conlangers use a random word generator. There is, though, such a thing as too random—even if you allow for phonotactic constraints. So while one doesn’t want to copy one’s own phonological biases, one also should plan for phonosemantic patterns.

Here, for example, is a common pattern in English: glitter, glimmer, glint, gleam, glisten…

You could add a few more in there, I’m sure, but this is enough to get the idea. In English, for whatever strange reason, a bunch of words that have to do with flashing lights start with a “gl” and have a high front vowel. There doesn’t seem to be anything crosslinguistic about this pattern (seems pretty random to me): it just exists.

If you’re using a random form generator, you will probably not copy over your native language phonological biases, but it’s also not going to produce any patterns like the “glitter/glimmer” pattern of English. Those patterns need to be handcrafted.

Another facet of language that’s hard to replicate with a random word generator (I appear to have switched topics. Going to stick with that…) is the uneven distribution of phonemes. For example, there are a ton of words in English that begin with /t/; not so many that begin with /θ/—and even fewer than begin with /ð/. It makes sense if you go way back in the history of English and move forward. But when one is standing on the other end and has a list of phonemes, it’s difficult not to just start each new word with a new phoneme, giving one an even number of words starting with each phoneme.

And that, of course, is just the beginning of the word.

I’m now no longer sure if I answered any question at all—or even if the original question has anything to do with what I said, or needed an answer. I’m pretty tired. And now I’m distracted because Erin’s watching the wedding episode of The Office. I need chocolate. And ice cream. Chocolate and ice cream! This I must have…

Tags: ,

4 Responses to “Lavaka”

  1. Ka kavaka Rejistania ti:

    A very insightful posting. Which is why I actually tried to hadcraft certain biases and structures in Rejistanian. These are hard to specify for me, but there are certain phonoaesthetics, which Rejistanis for example prefer for ‘bad’ words (‘bad’, ie: sejil itself is an exception, but jurutu, and tuku follow it for example).

  2. Ka kavaka Ember Nickel ti:

    Hmm, okay. I guess I’m just really willing to run the risk of the first and minimize the third, making up words according to whatever “sounds right” at the time. So I have about four times as many words starting with k than with v, but maybe they’re phonologically biased–I can live with that. (My word for “water”, for what it’s worth, is yūyē.)

  3. Ka kavaka Alex F ti:

    I dunno. I think that if your random word generator can’t do non-uniform frequency distributions (more [t] than [θ], etc.), it’s just thoughtlessly written, and you should ditch it (or tweak it) for one that can. There’s no sort of inherence to that limitation, no reason making a program do this is hard.

    On the other hand, I think that in any random word generator I’ve seen, you can achieve that effect even if the instructions don’t mention it, along these lines: instead of specifying say your C inventory as /p t k …/, specify it as /p t t t t t k k k …/, and you ought to get six times more /t/ than /p/, etcetera.

    But on the third hand, you sorta seem to be thinking of randomly generating a lot of forms and subselecting from them by hand. Then the selector’s own biasses are added into the mix, and it’s not (exclusively) a problem with word generators anymore.

  4. Ka kavaka David J. Peterson ti:

    But on the third hand, you sorta seem to be thinking of randomly generating a lot of forms and subselecting from them by hand.

    Well, of course. Doing anything less (or is it more…?) would be irresponsible.

    And while, of course, one’s biases factor into selecting words from a randomly generated set, I’ve found that it helps to give one ideas one might not have otherwise hit upon. I know I’ve had the experience, at least, where I look at a randomly generated word, and think, “Holy smoke! That is a licit word given my phonotactics!” It can be kind of fun. :)

Leave a Reply