Blog Archives

Relative Clauses in Dothraki

Today’s post is going to end up being rather long and grammar-heavy. Some readers will probably dig that, but it may be a bear for others, so by way of making amends, here is a picture of my cat Keli wearing sunglasses:

My cat Keli wearing sunglasses.

Click to enlarge.

From now on whenever you think of relative clauses, I hope you will think of the coolest cat on the planet (and that, of course, would be my cat).

Relative clauses are clauses which modify a noun. Here are some examples in English (with the relative clauses underlined):

  • The goat who loves me.
  • The octopus that I saw eating Twinkies.
  • The jaguar I gave a camera.
  • The penguin I saw Driving Miss Daisy with.
  • The duck whose uncle I tased at the Super Bowl.

In each case above, the modified noun plays some sort of role in the relative clause. Those roles are:

  • Subject: the goat is the one doing the loving.
  • Direct Object: the octopus is the one I saw.
  • Indirect Object: the jaguar is the one I gave a camera.
  • Object of a Preposition: in this case, the penguin is the object of “with”.
  • Possessor: the duck is the one that has the uncle I tased.

English is quite permissive in its relative clauses. You can relativize almost any role. Some languages are less permissive, allowing you to relativize only certain roles. In Kamakawi (a language of mine), for example, only subjects can be relativized (below, ana is “duck” and topu is “shout”; for more on relative clauses in Kamakawi, go here):

  • Ana poke topu i’i. “The duck that called me.”
  • Ana poke topu’u ti’i. “The duck that I called.” (lit. “The duck that was seen by me.”)
  • Ana poke topuku’u ti’i. “The duck for whom I shouted.” (lit. “The duck that was shouted for by me.”)

For Dothraki, I decided to “simplify” things (which, in all cases having to do with Dothraki, means “make things closer to English” [though English itself is far from simple]). I actually did this in several places on purpose. Dothraki is quite different from English, but it could have been differenter, if you’ll allow (though my spellchecker says that word is licit! How ’bout that…). I figured if people were going to try to use Dothraki, many would be at least familiar with English, so in order to make some rather foreign choices in some places, I dialed it back in others. The Dothraki relative clause is one of the places I dialed it back a bit.

In Dothraki, a bare noun of any case (in the embedded clause) can be the target of relativization, as can a noun that’s the object of certain (but not all) prepositions. To form a relative clause, one puts a relative pronoun whose animacy matches the animacy of the head noun directly after the head noun in the matrix clause. That relative pronoun takes on the case the head noun would display in the embedded clause. After that, one writes up the embedded sentence using VSO word order and with a gap where the relative pronoun would be.

That might be a bit much to digest, so let me illustrate with some examples. First, I’ll go through two simple examples: one where the target of relativization is a subject in the embedded clause, and another where it’s the direct object:

  • Adra fin tih anna. “The turtle who saw me.”
  • Adra fines tih anha. “The turtle whom I saw.”

The first sentence (where the turtle is the subject of the embedded clause) is pretty much identical to English; the second is a little different. Specifically, you’ll notice that the subject, which ordinarily precedes the verb in Dothraki, follows the main verb of the embedded clause. The reason it does so is the word order of embedded clauses reflects the older word order of Dothraki, which was VSO. This is often the case with natural languages (relative clauses are often a way to take a glimpse into the past).

In the case of Dothraki specifically, the pre-verbal position was, for many hundreds of years, a topic position. When the word order was VSO, a noun could be dragged up front in order to topicalize it. As time passed, the topic and subject positions became synonymous. In the case of relative clauses, though, since the relative pronoun occupies the topic position, the subject remains in situ. In matrix clauses in the modern language, you can actually get pre-subject topics (now that the pre-verbal position is the subject position), giving Dothraki stylistic word order variants of both OVS and OSV.

Hopefully that’s pretty clear. Now here are some other examples of relative clauses with nouns taking on different roles (note: some of the sentences below will require you to believe that a turtle could own and perhaps wield an arakh):

  • Adra fini tih anha arakh. “The turtle whose arakh I saw.”
  • Adra finnaan azh anha arakh. “The turtle I gave an arakh to.”
  • Adra finnoon ahajanak anha. “The turtle I’m stronger than.”

That last example, in addition to showing the ablative of the relative pronoun, also shows that comparands can serve as targets of relativization (something which not all languages allow [partly depending on how they handle comparison]). Similar examples can be formed with inanimate nouns, but the inanimate form of the relative pronoun is used. Here’s the declension patterns of the animate and inanimate relative pronouns:

Case Animate Relative
Pronoun (SG/PL)
Relative Pronoun
Nominative fin fini fini
Accusative fines finis fin
Genitive fini fini fini
Allative finnaan finea finaan
Ablative finnoon finoa finoon

Lastly, I mentioned that the objects of certain prepositions could be relativized as well. This is true, but it involves a construction which is probably one of the least familiar aspects of regular Dothraki grammar (to English speakers). In Dothraki, many prepositions can be used pronominally (or adverbially… I’m not sure which it is, honestly. I think it’s pronominally) when the noun phrase it modifies is a pronominal argument already present in the discourse. That may sound complex, but actually the easiest way to illustrate how it works is with a relative clause.

Let’s start with a sentence like the following:

  • Adra zimeme mawizze ma fesoon. “The turtle distracted the rabbit with a carrot.”

That’s pretty straightforward. Now what if you wanted to say something about that carrot? Here’s what you would do:

  • Anha tih fes finoon zimeme adra mawizze memas. “I saw the carrot the turtle distracted the rabbit with.”

First, the relative pronoun is put into the case it would have been assigned had it been preposed by the preposition (in this case, since the preposition is ma, which assigns the ablative case, the relative pronoun is in the ablative). Next, the preposition is put into its pronominal (or adverbial) form and stands for the whole prepositional phrase.

In English, we actually do have vestiges of a system like this, but the language is, often, considered stilted or antiquated. For example, if the carrot were already under discussion, we might say, “And the turtle distracted the rabbit therewith”. In Dothraki, rather than fronting the preposition (e.g. “I saw the carrot with which the turtle distracted the rabbit”) or stranding it (as above), the preposition is left in situ and put into this special pronominal form.

As I mentioned, not all prepositions work like this. Here’s a list of Dothraki prepositions that do work this way:

Dothraki English Pronominal Form
ha for, from mehas
irge behind, after mirges
ki by, because of mekis
ma with memas
mra in, into, inside memras
oleth over, above molethas
she on top of, on, at meshes
torga under, underneath, below metorgas
vi between, among, through mevis

When a preposition that doesn’t fit this pattern is used, a pronoun coreferent with the relative pronoun must be put in what would otherwise be a gap. It doesn’t sound great, though, so it’s usually advisable to try to reword the clause, rather than to resort to using the copy pronoun. Here’s an example:

  • Adra finnoon drivo anha haji moon. “The turtle for whom I died.” (lit. “The turtle whom I died because of him.”)

It’s not nearly as bad in Dothraki as the literal translation above is in English, but it’s not as elegant as other relative clauses that use the pronominal form of a given preposition, so if possible, it’s to be avoided.

As a final note, the distinction between the animate and inanimate relative pronoun is likely on its way out of the language. The default is becoming the inanimate relative pronoun (whose declension paradigm is simpler), and using it with an animate head noun is no longer as ungrammatical sounding as it once was.

That is almost all you’d ever need to know about relative clauses in Dothraki. This post is quite long as is, so I won’t go into indefinite relative clauses (I’ll save that for another time).

Oh, but as one last note, the post-subject verbal modifiers come right after the relative pronoun, if present, e.g.:

  • Adra fini ray tih anha arakh. “The turtle whose arakh I already saw.”

Aside from the indefinite relatives, that should be just about everything. Thanks for reading!

Citation Form

I got a question from Hrakkar which begins:

This thread brings up a good question: What is ‘lexical form’ for Dothraki?

To read the full question, go here. Basically, I think there’s two questions here:

  1. What is the citation or dictionary form of a given Dothraki word?
  2. What is the bare stem in Dothraki?

I’ll try to answer both questions.

To begin to answer both of them, first I’ll go over how the language is built. There are many different ways to build a language (and by “build” here I don’t mean construct so much as build up, or flesh out), and I’ve used different methods for different languages. Two different methods can be illustrated by glancing at the dictionaries of two of my other languages: Kamakawi and Zhyler.

Kamakawi is a language that is largely isolating with some agglutination. There’s no stem-internal alternation, and its writing system is glyphic and isolating (somewhat akin to Egyptian hieroglyphs). As such, each word kind of stands on its own. There are relationships between words, of course, but since many word forms can be used as verbs, nouns or adjectives, listing words separately makes more sense than listing them together. Here’s a sample of a page from my Kamakawi dictionary:

Part of a page from the Kamakawi dictionary.

Click to enlarge.

As you can see, in Kamakawi’s dictionary a single word is used as the head of each entry, and related words that differ in form get a new entry.

Zhyler is quite different. The script is alphabetic (and was meant to approximate the appearance of Latin), and Zhyler words (both verbs and nouns) are built off of a number of noun classes. Consequently, a single root will have somewhere between 3 and 17 forms associated with it whose phonological form is predictable, and whose meaning is often partially predictable. Here’s a sample of the dictionary that’s about the same size as the Kamakawi sample:

A sample of the Zhyler dictionary.

Click to enlarge.

Each root, then, gets its own entry, and words derived from that root (usually via noun class suffix) is listed under that entry. The idea for this type of dictionary came from Arabic, whose dictionaries are ordered alphabetically by triconsonantal root (which, if you know Arabic, makes a lot of sense).

As I think I mentioned somewhere, Dothraki is built in the same way Zhyler is (I like to think of Dothraki as being run on a Zhyler engine). Even though the languages are radically different, I flesh Dothraki out in the same way I flesh Zhyler out: by starting with a root and deriving words from it.

The reason this works well for me is that even though Dothraki doesn’t have noun classes the way Zhyler does (Zhyler has 17; Dothraki nouns, rather, fall into one of two broad classes: animate or inanimate), separate word forms tend to look different from one another, and are built in unpredictable ways. This is on account of Dothraki’s “pseudo-classes”, as I like to call them. Final vowels in Dothraki often serve no function other than to distinguish words from one another (one can easily imagine them dropping off some time in the future, as many word-final central vowels did previously).

Here’s one quick example using the root em:

  • emat (v.A) to smile
  • eme (ni.A) smile

A smile is, undoubtedly, related to the verb “smile” in some logical way. The final vowel -e though doesn’t define a process that takes one from a verb to a noun that describes an instance of a particular verb: It’s just a vowel used with this particular root for that function. Here, for example, are two other roots where this pattern doesn’t hold. First, the root yanqo:

  • yanqolat (v.A) to gather, to collect
  • yanqokh (ni.A) collection

That latter is a particular collection of something, not the act of collecting something. The next root is gach:

  • gachat (v.A) to figure out, to solve
  • gache (ni.A) place, environs

Many of these final vowels for inanimate nouns, then, form these pseudo-classes that have nothing in common with each other other than form (though there are patterns that hold if one considers a subset of the lexicon). By grouping such words under a single root, one can see how a given root has been fleshed out, and a single word will often make more sense in the context of its root than outside that context.

Another reason grouping words together by root makes more sense for Dothraki is that often words are not derived from one another, but derived directly from the root. As such, related words may have definitions that don’t look anything alike. By grouping them under the root, it’s easy to see that, ultimately, they come from the same source.

In my response to Hrakkar‘s comment I said this was going to be quick, so I’d better wrap this up. So now that we’ve seen what the dictionary looks like, more or less, I can answer the first question I posed above. In Dothraki, the various word types have the following citation forms:

  • Nouns: nominative singular.
  • Adjectives: singular uninflected.
  • Verbs: infinitive.
  • Other: maximal form.

That latter really only applies to prepositions like ma which can appear as m’ if they occur before a word that begins with a vowel. Anyway, those are the citation forms for each word, but they don’t tell the whole story. It’s important that (in my dictionary, at least) words are listed with their associated roots. Consider the following verbs (in their infinitive forms):

  • hoyalat (v.A) to sing
  • indelat (v.A) to drink

One of these roots ends in a vowel; the other ends in l. Can you tell which is which? Absent of some other mechanism (like a hyphen or a period), there’s no way. However, if you know the root of hoyalat is hoyal and the root of indelat is inde, then by simply having the infinitive, one can fill out the rest of the verbal paradigm.

Aside from that, the reason I chose the infinitive as the citation form for verbs is that it’s fairly stable. In most cases, the singular past tense of a verb will be the simplest form of the verb, but it will often look like another word form (for example, haqe is an adjective which means “tired”; it’s also the past tense singular of the verb haqat, which means “to be tired”). For that reason, it makes more sense to use the infinitive which will (almost) always be unique.

Other languages, though, do things differently. In Arabic, for example, the citation form of the verb is always the third person singular masculine past tense. That may seem downright absurd unless you know what verbs in Arabic look like. Here’s a partial paradigm of kataba, “to write”:

Present Tense Past Tense
‘aktub “I write” naktub “we write” katabtuu “I wrote” katabnaa “we wrote”
taktub “you(m.) write” taktubuun “you( write” katabta “you(m.) wrote” katabtum “you( wrote”
taktubiin “you(f.) write” taktubna “you( write” katabti “you(f.) wrote” katabtunna “you( wrote”
yaktub “he writes” yaktubuun “they(m.) write” kataba “he wrote” katabuu “they(m.) wrote”
taktub “she writes” yaktubna “they(f.) write” katabat “she wrote” katabna “they(f.) wrote”

Bearing in mind that the non-finite forms for a verb in Arabic often look radically different going from verb to verb, the third person masculine singular past tense form (which, given Arabic’s writing system, is written with just the three consonants of the root) is the obvious choice for representing the verb—plus, that form (CaCaCa, where C stands for a consonant) doesn’t occur anywhere else in the language (say, as a noun). It was made for dictionaries.

For Dothraki (to finish up the discussion of verbs), if you don’t list the root, it’s probably best to set off the infinitive suffix from the root (e.g. and Since Hrakkar brought up the Dothraki vocabulary list on the wiki, though, one can achieve the same effect by also listing the past tense singular form of the verb. The reason is that the past tense will be either the bare root, or the bare root plus -e. By comparing the infinitive and the past tense, then, one will know for sure what the root is.

For nouns, in addition to knowing whether a noun is animate or inanimate, one will also need to know a couple pieces of extra information (for some nouns, at least). For inanimate nouns ending in a vowel, there are two broad classes which I call A and B. Class A nouns lose their final vowel in the accusative; class B nouns take an -e in the accusative. Certain other nouns will have a modified accusative form (so the accusative of tolorro, “bone”, is tolor).

And (a bit of new information), there are also a very small number of irregular animate nouns. These nouns all end in i (actually a vowel followed by i). These take vowel-final animate noun case endings in all cases, but in the accusative, the i becomes a y. One noun like this is mai, “mother” (so the nominative plural is maisi, but the accusative plural is mayes).

All right, this short response has gotten unruly and taken up much more space than I intended, so I will cut it off here. If you have questions about any of the above, feel free to ask in the comments. If you’ve read this far, you’re a real davrasok. Hajas!

The Header Script

I’ve gotten a few questions about the script in the header (the one in black below the word “Dothraki” in white), so I thought I’d address it, even though it doesn’t relate much to Dothraki itself.

First (in case there’s anyone who can’t make it out), it’s intended to say “Dothraki” in English (works if you kind of squint your eyes at it [especially when you get to the “a”]). Even so, the script was not intended to be a Roman font: it was created for a conlang.

Back in mid-2010, when I was working on season 1 of Game of Thrones, I was also participating in a group conlang creation project (the language was called Kenakoliku). It didn’t end up being successful (group projects are very difficult to maintain), but it was fun at the time. Since my favorite part of creating a language is creating its orthography, I devoted my energies to creating a possible orthography for Kenakoliku.

I initially fonted up five proposals (just one character for each): two instantiations of someone else’s hand-drawn proposal, and three of my own. They’re shown below:

Five orthographic proposals for Kenakoliku.

Click to enlarge.

I actually liked the one called “Curvy Glyphs” the best, with “Kadani B” coming in a close second (for some reason it reminded me of those block Greek letters you see on fraternity houses…), but most everyone else liked what I called “Halfsies”: an orthography where the consonant was on the bottom, and the vowel on the top. As a result, I filled the font out, creating possible consonant and vowel characters without assigning any values to them (so people on the board could choose which ones they liked best). Here’s the chart I came up with:

Full version of the Halfsies font.

Click to enlarge.

From this sample, everyone (mostly) settled on form-sound combinations they liked, and I produced a font, a sample of which is shown below:

A sample of the full Halfsies font.

Click to enlarge.

The font had some problems, and I mostly fixed them so that the font worked in my word processor of choice, but then it still had problems in other word processors, and then the language itself kind of wound down anyway, so I abandoned the project and the script. If you’d like, you can download what remains of the font and toy around with it here (.zip). I warn you, though: the ligatures may not work on your end, and I’m no longer maintaining it.

Anyway, several months later, Game of Thrones debuted, and I started up a LiveJournal account mainly to make comments on George R. R. Martin’s LiveJournal. In order to do so, though, I needed an icon. I didn’t want to use a picture of one of the actors from the show (or, even worse, a picture of me), and Dothraki didn’t have its own script, so I was in a quandary. I’d always liked the Halfsies font, though (I always referred to it in my head as the Swashbuckler’s Script), and one of the characters kind of looked like a “D”, so I made it my LJ icon:

The Dothraki D.

And that’s been my little Dothraki icon ever since.

When it came to making this blog, I found a theme I really liked (props to digitalnature!), but the font in the header looked way too plain (sans-serif?! Decidedly un-Dothraki!). And since I didn’t want to actually go in and mess with the CSS, I just created a background image with the Halfsies font. In order to get something that looked like English, I had to pick and choose characters (and alter some, using Photoshop), some of which you will have seen in the image above:

Characters used in the background image.

Click to enlarge.

And eventually I had it.

So that’s the story behind the script in the header. The look of it was, indeed, inspired by Devanāgarī, but the actual structure is a bit different. The font above can’t actually be used to write English (it’d need a lot of work for that), but maybe if someone’s interested they can take it and manipulate it.

I feel like I should have something related to Dothraki in here, since this is a Dothraki blog. Let’s see… If I had to give “Halfsies” a name in Dothraki it’d probably be Lirisirazo (a class B inanimate noun, as all noun-adjective compounds are that end in a vowel), which means something like “Bladed Writing”.

Also, for the curious, Dothraki is now up to 3,163 words: More words than Mr. Padre Tony Gwynn has hits, but still a ways to go to catch Pete Rose (and my guess is Dothraki will have double his number before he gets even a sniff of the Hall of Fame).