Long (or Doubled) Consonants

It’s been a little bit, but I’m back with the first non-first post. Before getting to that, I should mention that I knew beforehand there was going to be some lag time with getting this blog off the ground. I told the Dothraki.org folks I would be starting up a blog before I’d even settled on a site theme, and then I put out my first post knowing that I’d be completely unable to post anything else for at least a week. My apologies.

There are already a number of great questions/topics over at the Dothraki Forum, but I thought I’d start small and answer Tim’s question first, since it’s come up before. The question is:

How are doubled consonants pronounced differently than single consonants? How about situations where you have something like ssh vs sh?

First let me address the orthographic question. Since the Dothraki don’t have a writing system of their own, I came up with a romanization system to write the language. To make it easy on everyone, I decided to restrict myself to ASCII (aside from the texts I send to the actors, which have main stress marked with an acute accent). When one does that, one is forced to use either unorthodox characters or digraphs for sounds for which no single roman letter is used (in this case, in English). The relevant sounds in Dothraki are (romanized form followed by IPA):

  • th [θ]: like the “th” in English “math”
  • sh [ʃ]: like the “sh” in English “shout”
  • zh [ʒ]: like the “z” in English “azure”
  • ch [tʃ]: like the “tch” in English “watch”
  • kh [x]: like the “ch” in English “blech!”

If instead of these digraphs I’d chosen, for example, þ, š, ž, č and x, then romanizing a doubled consonant would be as trivial as romanizing any of the other doubled consonants. Due to some of the peculiarities of Dothraki phonology, though, I was able to represent the geminate versions of these digraphs simply by doubling the first consonant:

  • tth [θθ]
  • ssh [ʃʃ]
  • zzh [ʒʒ]
  • cch [ttʃ]
  • kkh [xx]

By “peculiarities”, I mean that the sequences that these second set of digraphs could also represent happen to be disallowed by Dothraki phonology. Specifically, homorganic oral stop+fricative clusters are disallowed. So, in the case of kkh, for example, it will always be [xx], and could never be [kx].

As for ssh and zzh, when two sibilants ([s], [z], [ʃ] or [ʒ]) occur next to one another, the former assimilates to the latter. So a potential cluster like /sz/ would become [zz], /zʃ/ would become [ʃʃ], etc. Thus, there could never be [sʃ], for example, meaning that ssh will always be pronounced [ʃʃ].

That turned out to be a longer explanation than I thought… Anyway, now to the main thrust of the question: how these are cats pronounced.

Doubled or geminate consonants are common throughout the world’s languages, but they happen not to be common in many Indo-European languages (among them: English, Spanish, French, German, Dutch…). Some of the many languages in which geminate (or doubled or long) consonants are distinctive are Japanese, Italian, Finnish, Hungarian, Russian, Arabic and Turkish. In these languages, you must be able to perceive and pronounce long consonants differently from short consonants, or you risk hearing or saying the wrong thing.

For a quick natural language example, let’s take a look at Arabic. In Arabic, the middle consonant of a number of verbal roots can be doubled to produce a kind of causative verb. The verb for “to write”, for example (in the masculine past tense), is kataba. By doubling the t, it becomes kattaba, which means “to dictate” (as in “to dictate a letter”). You can hear this pair of words below:

If the two don’t sound distinct enough, we actually do have “doubled” consonants in English, but they don’t distinguish single words from one another. For example: Imagine two women named “Ally” and “Sally”. Now put “Miss” in front of each one. The pair should sound something like this:

Notice how the [s] is longer in “Miss Sally”? Now just imagine that that length was able to distinguish words, and that’s what we have in Dothraki. Here are some sample pairs which are distinguished by the presence or absence of a geminate consonant. The pairs are: ata vs. atta; ara vs. arra; asa vs. assa; and ana vs. anna:

And now for a pair of Dothraki sentences where the only difference is a doubled consonant:

  1. Anha risse jeloon. “I cut into the lemon.”
  2. Anha risse jelloon. “I cut into the cheese.”

As for how to actually produce these differences, it kind of feels different for each type of consonant. For stops (tt, kk, qq, dd, gg, jj, cch), it kind of feels like you pronounce the first one, and then hold your breath for half a second: everything in your mouth is still for a beat, and then you release the consonant. For others (fricatives like ss and zz, or nasal sounds like nn, or even liquids like ll), you allow the sound to continue longer than you would ordinarily. To me, the latter are a bit easier than the former, but with a little practice, it’s not too hard to get the hang of both.

Thanks for the question, Tim! I hope this explanation serves.


  1. Very nice.

    There seems to be some issue with the sound files though. When I play them in Firefox there is no sound and in Chromium they don’t start at all. It might be some addon that’s stopping them on my side though. The addon Netvideohunter picked them up though so I could hear them through that.

    Also could you give the nominative for lemon and cheese? I’m guessing they might have a final vowel that gets dropped in the ablative?

  2. It seems like the audioissue has to do with volume. At first they are all silent but when I click on the volume slider on the player ever so slightly the sound appears.

    1. They all work fine for me: play right when pressed and play at full volume. I’m using Chrome on OS X.

      For the two nouns: jela is “lemon”, and jelli is “cheese” (but the accusative is jel).

      1. The browser it doesn’t work on is Chromium for Ubuntu. I guess there might be a codex issue on my side with that browser. I guess we’ll hear if anyone else has a problem.

      2. Are both accusatives “jel”? It seems like they are both inanimate given the ablative and if jelli also is jel as accusative then they look to have the same accusative.

  3. Great basic explanation of gemination. I need to get into using gemination more, never really done much with it. Looking forward to seeing more awesome Dothraki stuff talked about here.

  4. Cheers!
    It’s a nice bonus to learn a bit of phonotactics, too.
    On the audio examples a couple of the single consonants (t and s) sounded to me already doubled – and the respective long versions sounded just exaggeratedly long (I think our neighbour estonians have a third consonant length, but us finns have only two, so it wasn’t familiar). Sure, the phenomenon is probably mostly due to bad listening. It’s hard not to listen over-analytically; anything sounds weird out of context and intently repeatedly listened to. And the stress might throw me off. And of course the words might also be intentionally exaggerated a bit.. But might there be more interesting stuff at work here? Maybe dothraki have a different rhythm, a bit longer consonants than finns? Is this generally a variable between languages? What would be the crux of such variable, basic length of consonants vs. basic length of vowels?

    1. Actually, when it comes to natural languages, the length of geminates vs. singletons as well as something called VOT (voice onset timing) varies greatly. For example, in English, voiced stops ([b], [d] and [g]) are lightly voiced, as voicing usually begins at the time of closure for the stop. With other languages, voicing begins before the stop.

      So, to answer your question, yes, languages differ with respect to this feature (I believe Estonian short stops are, in fact, shorter than most, which is useful for their three different consonant lengths). In Dothraki, certainly the languages I’ve learned are going to influence my pronunciation, and the language where I’ve had the most practice with geminates is Arabic (I believe they exist in Russian too, but they’re most noticeable when learning Arabic). I wouldn’t be surprised to learn that Dothraki geminates share more in common with Arabic geminates than other languages with geminates.

  5. (Sorry for not responding right away– it has been a bizarre week here in Reno) Thank you for the detailed explanation of the pronunciation, as well as helping me understand geminates a little better. It now appears I have some serious pronunciation practice to work on! (pronunciation has been my primary thrust so far, until I have some free time to work on starting to understand the grammar and syntax). One for certain I have to work on is the kk in Hrakkar!

    On jeloon, Am I hearing this right as je-lo-on and jelloon as je-looon ?

      1. Sounds jel-lo-on to me. How much do you, David, define the language by your own speaking? For most of us you are the gatekeeper of the Truth of Dothraki. But do you have a definite take on the canonical ideal Dothraki? Is you own pronunciation a measure stick by which to decide, what is right, or is there an image of ideal dothraki in your head that you just approximate the best you can? … Something else?

        1. The way I see it, I have a very particular idea about how at least one variety of Dothraki is spoken—and there was no chance of it ever being spoken on screen, for example, because I didn’t bother to describe it fully (for example, I pretty regularly turn /o/ into [ɤ] when it follows a dental consonant—by design—but a detail like that isn’t worth trying to explain, as pronouncing it as [o] still conveys the phoneme well enough). Due to the nature of the Dothraki themselves (long periods of time spent in small groups a part from other speakers; shorter periods spent together with others), though, it seems likely that different dialects of Dothraki would exist.

          Actually, this might make a good blog post… For now, though, I think there would certainly be different varieties of spoken Dothraki which would exist concurrently on the Dothraki Sea. Structurally, they’d probably be quite similar (they do still need to communicate, on occasion), but pronunciation may differ within a narrow set of more or less well-defined parameters.

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.