The Art of Language Invention, Episode 31: The Sonority Hierarchy

[00:00:00]In izumi, my name is David Peterson and this is the art of language invention episode 31, the sonority hierarchy.

[00:00:11]Now, the sonority hierarchy is a theory within linguistics that states roughly that more sonorous sounds are more likely to be nuclei than less sonorous sounds. For example, let's take these two segments here and say that this is a syllable. If this is a syllable, then a is going to be the nucleus and k is not going to be the nucleus because a is more sonorous than k. Uh, the sonority hierarchy lists types of segments roughly like this with the lightest color sound being the most sonorous and the and the darkest one being the least sonorous. So, vowels are more sonorous than approximants, approximants are more sonorous than nasals, fric- uh, nasals are more sonorous than fricatives, fricatives are more sonorous than affricates, and affricates are more sonorous than stops. So, roughly all of these, uh, segments right here, all of these lovable fellows, so for example, something like u is always going to be a better nucleus than something like z, for example. Uh, but z is going to be better than uh.

[00:01:11]And also, uh, there's even rankings within a group. So, within a group, the more the vocal folds vibrate during the production of a segment, the more sonorous that segment is. So, for example, z is a better nucleus than s in according to the uh, sonority hierarchy.

[00:01:31]Anyway, uh, this basically, if you look at this from the top down and kind of rotate it 90°, this describes a pretty good, uh, back half of the syllable. So, the vowel is going to be the nucleus and then everything else is, uh, less and less likely to be the nucleus, more likely to be a coda. You can flip that around and get the other half of the syllable and get a lovely, say, beginning of a syllable that looks like this. Uh snya, for example. And then flip that right around and get the back half of the syllable. einstk And you could get one single word that was something like nyeingk Great word.

[00:02:14]Anyway, um the sonority hierarchy is something that's supposed to hold across languages, and you can see it at work in many many many different languages. For example, we have this word for bread in Spanish, pan. And you can see that it lines right up with what is supposed to be the nucleus and what's supposed to be on the periphery. In German, we have Knopf, button, and it just lines right up again.

[00:02:37]Russian, tsvyet perfect, lovely. Arabic, qalb, heart. Uh Hungarian, könyv. And if you've never encountered this word, that indeed does come from Russian, kniga.

[00:02:48]That is indeed könyv, where it ends with a palatal nasal followed by v.

[00:02:55]Also has the vowel ü in it, which is not present in Russian, but whatever. Uh anyway, könyv works. Uh something like blind in English works very very well, as well as the English word spend, which works very well for the word Wait a minute.

[00:03:10]Spend isn't how you say spend, you say spend.

[00:03:14]Huh.

[00:03:15]Well, I guess that doesn't work. It doesn't uh It looks like the sonority hierarchy kind of fell down on the job here. In fact, take a look at these two words from Georgian. We have tma, which is the Georgian word for hair. Great, perfect, no problems with that. And then you also have this word, mta, a Georgian word for mountain.

[00:03:34]Huh. So, well, I guess the sonority hierarchy didn't work. And so that means it is basically a theory that's broken, and we should just toss it out the window, right? Well, uh not necessarily, because this also doesn't follow the sonority hierarchy, right? mkhli Um wait, wait. No, no, no, wait. No, this does follow the sonority hierarchy, right?

[00:03:56]Um, but uh, is clearly not a word of English, okay?

[00:04:00]But, this too does follow the sonority hierarchy and is a word of English.

[00:04:06]This may not follow the sonority hierarchy. We have spat here, uh, which is a good word of English, but does that mean that you just throw all the rules out and you can ignore it all for English? No, otherwise, you could, for example, get something like spat, not a good word of English, pat, not a good word of English. Both of those are fricatives, both of those are voiceless fricatives, but for some reason spat is okay, but fpat is not okay. Meanwhile, on the other side, pats is great for both.

[00:04:37]Patf is not.

[00:04:39]Patf is terrible.

[00:04:41]Um, and so, even though you can have pats and past, and you can have spat, but not sat, there are certain things that you can and cannot do even in a language that kind of deviates from the sonority hierarchy. And the point is this, the deviations from the sonority hierarchy are generally principled. Um, and basically, that's kind of what makes this thing really interesting uh, for a conlanger. Not that it holds across all languages all the time, but that it generally kind of makes kind of good predictions a lot of the time for many languages. Uh, if you remember in the Art of Language Invention, the book, I suggested that you kind of could use this consonant inventory as the starting point in a conlang because uh, as I put it, most languages have all of these sounds and all languages have most of these sounds. It isn't the case that all languages have all of these sounds, certainly not. Um, or even that all languages have most of these sounds, but it's a pretty good place to start. It's a pretty good spread and it'll set you like on good on good footing even though you're not going to say every single language on Earth must have all these sounds. Same with the sonority hierarchy. Not every single language obeys it in exactly the same way, but most of most languages obey most of the principles most of the time. And that's kind of interesting. So and it's also potentially could be somewhat predictive. So for example, we have this author this French author in French it's something like uh Jean-Paul Sartre, right?

[00:06:18]And you would pronounce it something like this and again this is the slashes because of course they're probably capital X's or whatever. But Sartre, something like that.

[00:06:28]And in English, what do we do with that?

[00:06:30]Because obviously we don't have those R's, sure. But even so, even if you replace that with our own R, you can't do that in English. You can't put an R on either side of a T like that.

[00:06:40]And so what happens? Well, we do something to repair it basically because of the sonority hierarchy or you know, something along those lines.

[00:06:50]So we start with maybe Sa- Sartre, something like that. We're like, well, we need to fix this up. So let's throw a vowel on the end. So we get Sartra, right? Um another way of pronouncing this guy's name is Sartre. We just pretend the R isn't even there. By the way, Brett Favre becoming Brett Farve, same type of thing. We saw something that we couldn't do in English and so we fixed it to fit our better understanding of what a syllable uh structure is. And that is at least partially predictable by the sonority hierarchy. For example, like we allow bets and best even though one of them violates the sonority hierarchy theoretically uh technically.

[00:07:33]Um but we will allow Bart and not um Botre.

[00:07:37]Botre. We just won't do that. There are just certain things that that we won't do. Anyway, the entire point of this is that just because you have a theory that doesn't work 100% of the time for all languages, doesn't mean that it can't be a useful tool when you are building your own conlang. It can be a nice place to start, and then after that you just describe where the deviations occur. Uh but it's better than kind of a blank page, you know? It It's rather than saying, you know, anything can happen in this language at any time. It's more like most of the time this follows the principles of the sonority hierarchy, except for these instances, in which case the following happens. Anyway, just throwing that out there as a general way to treat uh theories, predictive theories like this from linguistics when it comes to creating your own language.

[00:08:28]Anyway, if you have a question that you'd like to see answered on the show, feel free to email me at [email protected].

[00:08:35]Otherwise, I will see you next time.