Post by Janalyn Guo

David J. Peterson is a writer and language creator ("conlanger"). He began creating languages in 2000 while attending UC Berkeley. After getting his master's degree at UC San Diego, he went on to create languages for HBO's Game of Thrones (2011), Marvel's Thor: The Dark World (2013), Syfy's Defiance (2013), the CW's Star-Crossed (2014), and Syfy's Dominion (2014). In 2007 he helped found the Language Creation Society.

Janalyn Guo: What role did your enjoyment of literature and storytelling play in your initial interest in conlanging?

David Peterson: Whether one intends it to be or not, the lexicon of a language is the story of its speakers. Though all languages are mutually translatable, the specific set of words a language has—their etymologies, their interrelationships—is unique, and is a product of that language’s unique history. As a language creator, one has the responsibility of shaping that history. In effect, it’s like creating a character or a setting—like Macondo or Yoknapatawpha County. The difference is that rather than the people and events taking center stage with the language filling in around, the language is the main attraction—with people and places often hinted at or alluded to.

JG: How many language systems have you created so far? Which are your favorites among 1) the languages you've had a hand in creating, 2) invented languages made by other conlangers, and 3) the languages used widely today?

DP: I’ve started at least 27 language projects that have had enough thought behind them to have a name. They’re not all equal in quality or substance, though. Some have barely 100 words and their grammars are poorly contrived. All conlangers go through this, though. We get better as we go along.

(1) Irathient is probably my favorite, if I have a favorite. Either that or Kamakawi, which is my largest language outside of the ones I’ve created for shows in recent years. With Irathient, though, I just did whatever I thought would be the most fun. It was fun to create, and fun to use (though difficult).

(2) I have a lot of favorites. Some that always come to mind are Doug Ball’s Skerre, John Quijada’s Ithkuil, Matt Pearson’s Okuna, Denis Moskowitz’s Rikchik and Sylvia Sotomayor’s Kēlen. Also, even though it’s early yet, Sylvia’s new language, Sodna-leni, is brilliant. I hesitate to say that because it’s so early, but every time I look at it, I’m really blown away. Plus, Sylvia’s someone who’s only ever worked on one language her entire life; that she even sat down to create a new one is news. It’s still growing, but it keeps getting better. Her ability floors me.

(3) You said languages, not created languages, so I’m going to assume you meant all languages. Hawaiian is my favorite language, without a doubt, but I also admire the beauty of Arabic’s grammar. Having said that, though, you learn Modern Standard Arabic in school, which isn’t what people really use in day-to-day situations, so it is a bit of a construct. Even so, of all the Semitic languages, I think Arabic really hit it out of the park—and the same is true of Hawaiian and the Polynesian languages.

JG: What was particularly fun and playful about the creation of Irathient?

DP: Well, to me a noun class system is basically a game. The first time I learned about Swahili, it sounded like just about the coolest thing in the world to choose different nouns and an adjective to see how the words changed: kisu kikali (“sharp knife”), visu vikali (“sharp knives”), mgeni mkali (“sharp stranger”), wageni wakali (“sharp strangers”), etc. I wanted to replicate that with Irathient, since I hadn’t done that in a conlang, except in a few early ones that aren’t noteworthy. If you don’t have the pressure of having to make yourself understood, working with the subsystems of language is just fun!

JG: I like what you said earlier about how conlanging is like creating a system that hints at and alludes to people and places. The invented languages that you mentioned above -- Kamakawi, Skerre, Ithkuil, Okuna, Rikchik, Kēlen -- what kind of speakers do they evoke for you? Is the strong evocation of a hinted-at reality an indication of a conlanger’s skill?

DP: Well, the question is quite amusing, depending on the language you’re talking about. For example, Denis Moskowitz’s Rikchik is spoken by rikchiks: six-foot-tall green aliens with one gigantic eyeball, no mouth, and 49 tentacles. It has no sounds: it’s signed using 7 of these 49 tentacles. So the language does quite necessarily evoke its “speakers” rather strongly.

Others are different. For example, John Quijada’s Ithkuil isn’t intended for a fictional culture: it’s intended for our use. Consequently, the culture it reflects is our own: the real world. And the nature of the project is to use the language to speak as unambiguously as possible using as few linguistic tokens as possible, so it kind of eschews the vagueness of natural language. It’s an inherently different kind of project which we call an engelang (an engineered language), and consequently evaluating by the same standards one would evaluate a language like Kēlen simply wouldn’t make sense.

But as for the others, yes, in a way. That is, there are two types of skills when it comes to conlanging: the nuts and bolts of the language, and the artistry of the lexicon. Those who are good with the latter, or who have a very strong sense of their speakers, will, perforce, evoke their speakers in the words they choose. Those who are not will not. I was one who wasn’t when I started out. It took a lot of learning on my part to get to where I am. Others like Sylvia and Sally Caves, whom I haven’t yet mentioned, were naturals, and I’m always stunned by the ingenuity and vividness of their invention.

JG: At what point in ones experimentation with conlanging is a language born?

DP: There’s no real cutoff point. The same is true of the question when are two dialects distinct enough to become separate languages: at some point, it just happens. There are certain key moments in the development of any language, though, and those usually serve as a guide. For example, there’s a point where the grammar is mostly done, aside from the grammar that may be introduced by a key lexical item here and there (e.g., a verb that takes a subordinate clause structure that hasn’t been used prior). The language may not have a large number of words, but if one can bring a language to a point where any potential sentence could be translated if there were enough lexical material, that’s a defining moment. After that, it’s usually different points in the lexicon’s size (100 words, 500 words, 1,000 words, 2,000 words, etc.). Once there’s a system in place that can handle translation, there is a language. It may need more words and may have no speakers, but the system is the language.

JG: I know that some conlangers explore the ways in which language patterns reflect thought patterns, language mirroring pure thought. How achievable do you think this is?

DP: Personally I don’t think we understand enough about thought to be able to encode it directly in language. The attempts have value, though. Anything that allows us to look at something old in a new way has the potential to trigger some sort of epiphany. I feel like we’re a long way off from being able to understand actual thought in concrete terms—and also feel the answer will probably come to us from neuroscience, if we ever get it.

JG: Do you think a computer could create a language for humans to use with each other? I’m curious about how AI might influence language creation, and vice versa.  

DP: I have a different opinion about AI now from what I did when I was growing up in the 80s with all the “computers ruling society” horror/sci-fi movies. Computers are much more powerful than we imagined, but there’s no there there. A computer could certainly automate most of the tasks involved in language creation, but there’s no art there. The art would arise in its use. It would take a user to find the beauty in the artificial construct and start using it and selling it; there could be no intentionality. Of course, we’re a long way off from that. It’s not as if we don’t have the technology: there’s just no interest. There isn’t a crack team of programmers trying to create the perfect language creation algorithm, because there’s no demand and probably no money in it. There are a lot of other problems they’ll be tackling before they get to conlanging.

JG: I enjoyed your TED talk and was pretty astounded by the metaphorical quality of the Dothraki language (for example, the Dothraki origins for the words “to dream” and “tree”). Each word is like a little chest to unlock, and there’s a lovely poetics involved. What gave you the idea to mold the language in this way?

DP: First, time is important. It’s easier to do better work when you have the time to do it. After that, though, it’s kind of born of necessity and constraint. In crafting a lexicon, you start out with nothing and very slowly edge out into the vast enormity of describable experience. Ultimately, a language will need to be able to discuss everything, whether it does so with a single word or a series of words. When I come to the next word or concept I have to describe—whether it be because I need it for a script, or because that’s where I want to go next—there are a number of practical questions (should this word be a compound, a new root, a borrowing, etc.), but the most important question is: How do I want to realize these concepts for this group of speakers? And there’s no reason that the answer has to be with a completely new word form every time. If it were, then, yes, a computer could do it quite efficiently—could probably produce a million word lexicon in a matter of minutes. The result, though, wouldn’t be of any practical interest.

JG: What elements of Dothraki culture and ideals directed your creation of their language? And similarly, what Valyrian ideals directed your creation of the High Valyrian language?

DP: With Dothraki, there are a couple of cultural details we get of them which say a lot about their culture. For example, their cultural fear of the sea gives you an idea about what kind of experience they have with the land they live on, and an idea about where they will and won’t draw their metaphors from. Additionally George R. R. Martin notes several times that Dothraki do everything of importance under the open sky. This proved fruitful for me in creating the metaphorical framework I needed to discuss things that are good and bad. For example, in English we have a high-low metaphor that’s fairly common (“that’s low,” “underhanded,” “high-minded,” “on the up and up,” etc.). In Dothraki this is replaced with a metaphor contrasting things which are concealed or covered with those that are not, and this metaphor manifests itself in various ways in regular speech, as the height metaphor does in English.

The Valyrians were (and continue to be) much more difficult since we know virtually nothing about them. We know they had dragons and that they were a technological, cultural and martial power, and that their civilization was mysteriously destroyed, but we know little else. For that reason I’ve shied away from doing any heavily culture-inspired work for Valyrian. The distinction of High and Low Valyrian, though, was a nice one, as I borrowed the height metaphor into Valyrian (similar to English but contrasting with Dothraki).

JG: In an interview with Wired, you described the Dothraki vocabulary you invented as being "entirely à priori." What would be the template for an à priori language?

DP: First, I should have written “a priori” (the term is Latin, not French. My mistake!), but this term—and its sister, a posteriori—has a special meaning in conlanging. An a priori conlang is one whose lexical material is created whole cloth. An a posteriori conlang is one whose lexical material is based on another language. For example, Andrew Smith’s Brithenig is a language that presupposes Latin took over in Great Britain rather than the native Celtic languages. Consequently, instead of Welsh, Gaelic, Cornish, etc. being spoken, everyone spoke Latin. Brithenig, then, derives its vocabulary primarily from Latin, but filtered through the native languages, producing new Romance languages that quite a bit like Celtic languages. In order for the experiment to make sense, the vocabulary must be drawn from existing languages. With a project like Dothraki, though, it wouldn’t make sense for there to be words from our world in it, since the universes are distinct.

With that understanding of the terms, you can probably see that there is no possible template for an a priori language: it’s whatever the creator can imagine. A priori languages are as diverse as Dothraki and Timothy Ingen Housz’s Elephant’s Memory, which is entirely pictorial.

JG: Is it challenging to imagine past this world, to sort of unhear the sounds we associate with this universe to create these languages?

DP: Yes, this is one of the first and most important lessons every conlanger must learn. The best way to do it is simply to expose oneself to as many languages as possible. That is, one won’t know if something one’s language does is uncommon unless one learns other languages. Our first language becomes our entire world, and how could it be any other way? Learning as much as one can about other real-world languages and cultures is paramount. It frees one from the bonds of one’s mother tongue.

JG: In what ways has it been different designing languages for shows versus designing languages on your own?

DP: The major difference is the presence of deadlines. Working on your own language, you make your own deadlines, and have the freedom to make mistakes which can later be corrected. A show’s airdate, though, is an absolute deadline, and mistakes that make it into the show are there forever. That is unfortunate, since the time allotted is not sufficient. Mistakes do make their way into the show, and it’s disappointing, to say the least. If conlanging is an artform (and I’m certain it is), the best work will not be done for a production: it will be done on the creator’s own time. As for the latter, even if I stopped working for shows and movies right this moment, I’d continue to conlang for the rest of my life. I don’t know if it’s a question of motivation so much as compulsion. It’s what I do.

JG: It seems like an authentic language is often a missing dimension from certain books or movies with sci-fi, futuristic, or fantastical premises. In my mind, a unique dialect and language could be just as important as the bizarre physical forms of the invented world, as authenticating details. Are there sounds that we associate with the past, with the future, and with the strange?

DP: There are certain impressions we have (and by “we,” I refer to English speakers) when it comes to foreignness and alienness. They are impressions, though, and are entirely subjective. I’m happiest when I work on a project where the producers/directors/writers understand that it’s not “weird” sounds that make something alien or “harsh” sounds that make something foreign: it’s the character of the entire language and culture itself. Both culture and language grow organically, and it’s the natural evolution in an unnatural, alien or foreign place that will, of necessity, produce that sense of alienness or foreignness. It shouldn’t be faked.

JG: How does language creation seep into your other writing?

DP: I think much more about each word I’m using than I did before I started creating languages. It’s really changed my relationship with English, which is bizarre. And there’s no going back. Once you’ve crossed that bridge, you’re done—and it can be a blessing and a curse. For example, noting just how many linguistic possibilities there are when it comes to reifying any particular construction, I’ve started to lose my grasp on just which preposition is appropriate where. I can usually come up with an argument for any of them in English.

I wouldn’t trade it for the world, though.

JG: How might one construct an aesthetics for a written language; what might be the considerations?

DP: This question could be answered with a book. In effect, most conlangs are written languages, in that it’s rare that they’ll pick up speakers. Consequently they exist (to the extent that a language can extent) in roughly the same state as dead languages do. However perhaps you meant orthography…? If indeed you meant exclusively written languages, I encourage you to take a look at Elephant’s Memory, which I alluded to above, and Rikchik. Neither are speakable, and so are exclusively written. I tried my hand at a pictorial/iconographic [language] in 2005. I’m not too happy with it; it’s been abandoned at this point. Blissymbolics was probably the first such attempt (or intentional, serious attempt) in history. Either way, when you’re talking about any language and evaluating a language, the first step is to ascertain the goals of the creator and judge the work based on those goals. For example, with a different frame of reference, one could look at Brithenig and say it was totally unoriginal, since all of its vocabulary comes from other languages. Creating original stems, though, wasn’t the point of the project, and so one couldn’t judge the language based on that criterion—but one should for a priori languages. When it comes to a written form, there are a number of questions one has to ask before evaluating—for example, did the system evolve naturally (as did the Roman alphabet), or was it a construct (e.g. the Cherokee syllabary)? What level of technology is the society at? How was it written—using what implements? Once these questions are answered, one can evaluate the system based on its merits—and once that’s done, then based on the overall impression it gives. That would be the place to start.

JG: Maybe to tie these concepts together, I'd love it if you could translate a phrase from English into a few of your invented languages and walk our readers through some aspects of those languages.

DP: This is kind of a contrived example, but it’ll illustrate a commonly used but usually interesting concept. In English: "The father believed his daughter." In English, “believe” ultimately derives from the same word that gave us “love,” with the connection between the two concepts probably being something like the modern expression “hold dear.” Here’s how it comes out in some of the languages I’ve created.


Ave shillo ohar mae.

Father believed daughter his.

In Dothraki, the verb shillolat derives from shillat which is related to shilat, “to know.” Certain verb stems at a time in the past would double their final consonant to produce a verb of greater duration or impact than the original verb. So while shilat is “to know,” shillat is “to trust” (i.e. to know something several times over or for an extended period of time, and so it can be trusted). The suffix -o adds focus to the process, so that “to believe” is more “to come to trust over a period of time.” Consequently, the word shillolat carries with it associations with tried and tested knowledge (i.e. you believe someone because you have known them to be true several times in the past).


Kepa zȳhe tale pāsiles.

Father his daughter believed.

In High Valyrian, the word pāsagon means both “to trust” and “to believe.” I went back and forth on precisely which tense to use for the verb, because the usual past tense is the perfect. I went with the imperfect here, because the perfect gives us more of a telic reading, which would mean that the most natural interpretation would be “The father trusted his daughter.” Apart from context, using the imperfect would more easily guarantee a reading of “believe,” but it would assume some sort of ongoing context. Consequently, it wouldn’t feel as definite as the English (i.e. it would be a simple statement of fact). To get the definite reading, it would require the verb to be conjugated as pāstas, but without context, it would be more likely to be understood as “The father trusted his daughter.”


Zezik abishi zbaba zwinyazwa.

Her-he-did believe father daughter-his.
A lot of z’s in the Irathient because the father and daughter are of the first noun class reserved for Irathients. The theme consonant of the first class is z. In this case, the word for “believe” derives from the same root used for possession. In fact, used with a different auxiliary, it would mean “to hold” (i.e. if you replace zezik with zezi, it’d actually mean “The father is held by his daughter”). In Irathient, auxiliaries and verbs are combined to get unique meanings, so the words wear their etymologies on their sleeves.


Tande do tave re finjila.

Father sbj. daughter obj. believed.
Castithan rarely uses pronouns if it can be avoided—and in this case, the use of the familiar term for “father” suggests that the daughter is his own. The etymology is somewhat similar Dothraki’s, in that the word is derived from the word hinjilu, which means “to know.” In its oldest form, this word meant “to see.” In deriving finjilu, though, the honorific/augmentative prefix fa- was added, ultimately fusing with the stem and deleting the h. Thus, believing is a kind of refined form of knowing—that is knowing without the need for overt proof, on account of the circumstantial evidence.


Abor adyan difera yaya.

Father believed from-daughter his.
Sondiv is another language which derives “believe” from a possession verb. In this case, the verb idi means “to hold.” By adding the -m suffix (which shows up as -n at the end of a word) it produces a word that is metaphorically related to the original in a tangential way (other pairs of words derived in this way: sobuviv “stillness” ~ sobvumiv “safety”; irus “to touch” ~ irzon “to experience”; ison “to give” ~ ismon “to help”). The object associated with the new verb, though, can only be an idea or a belief. To make this sentence work, the verb is used intransitively, and the ablative clitic di- is used with “daughter.”


Awnoh vawna tatcha memaj.

Father his daughter believed.
Væyne Zaanics is a language I created with Nina Post for her novel The Zaanics Deceit. In the book, two families from the 14th century create a language that they pass on to their children, so even with the fictional history of the book, the language is a created language, as opposed to a natural language. All words are divided according to whether they are divine or earthly, in the old sense. Thus the word awnoh is the word for father, but its counterpart, Æyneh, is the word for God. Similarly, memaj means “believed,” but memæj would mean “knew.” The idea behind this pairing is that certain knowledge is divine; belief without knowledge is earthly.


Ka haleke fala ie laya tilea.

Did believe father the daughter of-him.
In Kamakawi, the word hale means “to think” or “to opine.” By adding the applicative suffix -ke, one can promote a prepositional argument to direct object position. This is similar to Sondiv, which requires a preposition for the thing believed if it’s animate rather than using the regular direct object morphology, except that Kamakawi does it directly (Sondiv can too, actually, but it’s more natural in Kamakawi). The verb might actually end up having additional interpretations, so context would be needed to ensure the correct reading.

*   *   *

Other Invented Languages Mentioned in the Interview:

Skerre by Doug Ball

Ithkuil by John Quijada

Okuna by Matt Pearson 

Rikchik by Denis Moskowitz

Kēlen by Sylvia Sotomayor

Brithenig by Andrew Smith

Elephant’s Memory by Timothy Ingen Housz