How to start making a conlang?

If you're new to these arts, this is the place to ask "stupid" questions and get directions!
Post Reply
somerandomkid
rupestrian
rupestrian
Posts: 1
Joined: 03 Jan 2021 08:27

How to start making a conlang?

Post by somerandomkid »

I'm new to conlanging and worldbuilding in general and have no idea where or really how to start. Can anyone help? Sorry if this isn't the right place to post this.
User avatar
sangi39
moderator
moderator
Posts: 3024
Joined: 12 Aug 2010 01:53
Location: North Yorkshire, UK

Re: How to start making a conlang?

Post by sangi39 »

somerandomkid wrote: 03 Jan 2021 08:46 I'm new to conlanging and worldbuilding in general and have no idea where or really how to start. Can anyone help? Sorry if this isn't the right place to post this.
Honestly, I think this is one of those questions that has many different answers, and each person has their own method. However, I'll try to answer this with the way that's been my experience.

First step for me was reading, about real-world languages (natlangs) to start (this had already been a hobby, I just wanted to learn them, but kept flitting from one language to another), but I ended up wanting to throw everything at my prospective conlang so it ended up being what some call a "kitchen sink conlang", apparently a common thing for early conlangers, but not necessarily bad.

Then I found The Language Construction Kit by Mark Rosenfelder, which was basically where I went "ohhh, moderation", and it really kicked off my reading into how natlangs actually work. What they do, what they don't do, what various tendencies pop up (fairly good source for this sort of thing, honestly, is Wikipedia, WALS, and then following up any thoughts of "oo, that's interesting" with a quick run to Google or down to the links down at the bottom of Wikipedia). Oh, we also have our own resources thread as well, and asking question never hurts, and presenting what idea you have, in places like the Quick Questions threads.


As for an actual "process", I tend to go:

1) phoneme inventory
2) phonotactics
3) allophony
4) inflectional morphology
5) syntax
6) lexicon and derivation

That's not really a step by step thing. The more you work on a language, the more you'll find yourself going back and forth between different parts, reworking bits of the inflectional stuff while you work on syntax, maybe changing a part of the phonology, etc.



Other people will do things differently, and you might find that the way I work might not work for you (it barely works for me, but that's more down to spare time than anything else. After a point I can't do in-depth work), but something else will.


So, personally, I'd say the main things are to read, to have fun, and to find a process that works for you, as well as to ask questions and show off what you've got.

Really, the same goes for world-building. Don't be afraid to mess around with stuff as you work. Throw an idea at it, see if it sticks right, and if not, then either remould the idea until it does fit, or leave it to one side and throw something else at it [:)]
You can tell the same lie a thousand times,
But it never gets any more true,
So close your eyes once more and once more believe
That they all still believe in you.
Just one time.
User avatar
eldin raigmore
korean
korean
Posts: 6352
Joined: 14 Aug 2010 19:38
Location: SouthEast Michigan

Re: How to start making a conlang?

Post by eldin raigmore »

[+1] [tick]
What sangi39 just said.
[tick] [+1]
Salmoneus
MVP
MVP
Posts: 3030
Joined: 19 Sep 2011 19:37

Re: How to start making a conlang?

Post by Salmoneus »

There are a number of guides - most famously the LCK, which sangi has linked. I believe David Peterson (the guy who made Dothraki) also used to have a guide up on his site.


However, the first thing I'd say is: think about what you want out of this, or at least what you want first.

If you're just looking for some placenames for a story, then I'd suggest focusing your efforts on phonology (what sounds a language has) and in particular (and this if often overlooked!) phonotactics - how those sounds fit together. These are important in giving distinct flavours to names: you'd know instantly that Hape'oko'i and Tanekhawa'e were named by people speaking a very different language from the people who named Cheneaux and Valpáseran. Orthography (how you spell sounds in the Latin script) also matters when it comes to shaping reader expectations: Schänow, Chieneaux and Xsehnòe may all be intended to be pronounced the same way, but the choice of spelling convention helps shape the immediate assumptions readers make about them.

If you want a language for a novel, again your emphasis will probably be on superficial appearance. I'd personally suggest, in that case, taking a real language and borrowing most of the grammar, and then deciding what you want the words to look like (not from the donor language). This gives the sense of a real language (because the nuts-and-bolts are taken from one) without a huge amount of effort.

That said, effort can be needed for more complicated projects. Most of us who conlang for any length of time know quite a lot of linguistics, so you kind of have to be willing to look into that. That said, most of us learn along the way, rather than going out and studying.

One useful resource for people who don't know much about languages is something like WALS, the World Atlas of Language Structures, which is free online (at wals.info). It's basically a survey of languages, broken down into a hundred or so specific questions, ranging from simple things like "how many consonants are there?" through to more recondite issues like "how productive is the antipassive construction?" and "do hortatives supplete?" - for each topic, there's an article explaining the different possible answers, and a map showing the frequency of different answers in different areas. I don't recommend reading it from cover to cover, but its articles are, I think, a good source of inspiration, as they often show you how differently languages can handle things that you might not otherwise even realise could be handled differently.


If you do want something more substantial, two common approaches to an early conlanging are:
- picking a real language or language family and creating a new daughter language
- making a language that's intentionally very unlike your native language

The former gives you an immediate framework. Typically you'll evolve your vocabulary from that of the parent language through regular and plausible sound changes* - and other languages in the family can guide you as to what those may be - and have a grammar that is similar to other languages in the family, while still allowing you to change some details. This approach rewards a willingness to research - to see how sound changes, semantic drift (the tendency of words to change meaning, and hence have to be replaced by other words over time) and grammatical details work in related languages. On the other hand, it makes it easier to create a large vocabulary - there are online tools that let you directly, automatically apply your chosen sound changes to many words simultaneously - and it's the easiest way to create a plausible, solid language relatively easily. The "traditional" (not because people are told to do it but just because it's obvious) approach is the "romlang" - creating a pretend romance language. This works because there's a huge amount of data available about romance languages (French, Spanish, Italian, etc), and your finished product will naturally resonate with people who have some familiarity with romance languages and/or latin (i.e. everyone in Europe or America). But there's plenty of other possibilities.

The other approach is less disciplined, since you can do what you want. It often involves reading instead about the weirdest things that languages can do and trying to fit them into your language. Less research work is required, since there's no wrong answers. On the other hand, you don't have your hand held to the same degree, since you don't have an immediate guide to follow, and you have to make more decisions for yourself. Many people find it harder to satisfyingly create words out of nothing, and you're also unlikely to immediately have the same resonance - and potential for injokes - with words that don't mean anything to anybody. People also often find that the things they've stuffed into their language don't really 'fit' - people often start out with bizare, improbable, 'kitchen sink conlangs', and gradually refine them down to (or replace them with) less weird, more focused languages. Very often, these "a priori" (not based on a specific real-life language) languages end up nonetheless being loosely inspired by particular language families or areas of the world, so the line between a priori and a posteriori (based on an existing language) conlanging tends to narrow over time.


Most people experiment with both these approaches early on; some people settle on one approach or the other, while others continue to switch between them or try to blend them (eg, making an a priori language, but 'setting' it in a real place and having lots of loanwords and grammatical influence from real languages).


----------------


That all being said, I'll take a step back and say that there are multiple possible ideologies of conlanging.

On this board, and in my experience most online communities, but certainly not all, the emphasis is on what we call "naturalistic" conlangs: conlangs that look like they could be real languages. This is partly because we actively want to make conlangs that look real... but to be honest, it's also because of a lack of other systems of judgement: "how naturalistic is this?" is a simple way of judging someone else's work, and of judging and guiding your own work. It's not necessary... but without it, a lot of people find themselves stranded in "why should I do one thing rather than any other thing?" paralysis.

However, it's also possible to take other approaches. One big one is 'auxlanging' - creating a language either actually or hypothetically intended to be spoken by the whole world. Naturalism may be useful here, but isn't necessarily the primary consideration - instead things like "ease of learning", "lack of ambiguity", and "non-bias" may be more important. Auxlanging often gets looked down on here - the world has moved on, we've realised that these goals aren't as straightfoward as they seem, and unfortunately the people who are really into auxlanging tend to be a little bit... intense. [they pop up here a couple of times a year: "I've spent 20 years inventing this completely unique language that is immediately understandable to everybody in the world and will eliminate all poverty and war!!!"... yeah mate, just like the last 50 perfect-languages-that-will-solve-all-problems...] However, as long as you recognise your auxlang is never actually going to be adopted, it can still be an interesting exercise.

The other option is the "engelang" ('engineered language'). These are languages that don't try to look like normal human languages - or even actively try not to. Sometimes they have a particular philosophical goal, like minimum ambiguity, or maximum intelligibility to artificial intelligences; other times they're just more abstract "what might a totally weird alien speak?" thought experiments. People have less to say about them, because they intentionally sidestep the usual "languages don't look like that, that doesn't look realistic" criticisms - and sometimes the creators are philosophically naive. But again, it can be an interesting intellectual exercise.


And then there's the type of conlang that most outsiders mostly see: the commercial conlang. Weirdly, relatively few actual conlangers are interested in making languages like this. These languages are primarily about creating particular impressions in the audience, and they don't absolutely care about 'naturalism' if it gets in the way - either for logical reasons (Tolkien's elvish languages - Quenya and Sindarin - look quite naturalistic, but their evolution is intentionally non-naturalistic, because the speakers aren't humans) or simply out of laziness (why bother making it perfectly plausible if nobody in your audience can tell the difference?) or concern for ease (if you want fans to learn it, don't make it too complicated!). However, these languages often do have a lot of naturalistic details behind the scenes, partly because the people who make them usually care more about languages than their audiences do. So, although Klingon's core project is just "sound violent and alien and be intentionally weird and alien to English speakers", Okrand stuffed a lot of details from native north american languages into it, because that's what he enjoyed. Quenya and Sindarin intentionally evoke Finnish and Welsh aesthetics, and their grammars display Tolkien's philological learning - in particular, the real art may be the simulation of philology itself (Tolkien often wrote about his languages as though he were a researcher rather than a creator, and in a way his languages' evolutions are intended to work in the over-simplified, under-understood way that historical-linguistic reconstructions work, as it were).



----------------------


* if you want to do naturalistic conlanging, then "sound change" is perhaps the most important concepts: words evolve over time, not randomly, but through the application of completely regular, exceptionless, easily-stated "sound changes". So English has "father" where Latin has "pater", because the ancestor of English has gone through two sound changes in which p>f and t>th (in this case the consonants of the shared parent language, Proto-Indo-European, are unchanged in Latin, though Latin of course went through its own changes). This convenient principle isn't actually true - real sound changes are not completely regular, do have exceptions, and can occasionally be very complicated to describe. However, it is true almost all of the time, so should be the general default assumption. A lot of other changes in languages can be blamed on sound change - for instance, Old English lost a lot of its final syllables through regular sound change, which resulted in the loss of noun cases, which forced the development of modern English's stricter word order. [but sound change can't be blamed for everything]
User avatar
Torco
sinic
sinic
Posts: 303
Joined: 14 Oct 2010 08:36

Re: How to start making a conlang?

Post by Torco »

I think what Sal calls a commercial conlang is kind of the same thing as naturalistic artlangs: even if we say the intent is different, is it though? those little easter eggs in klingon, well, they sure seem like artistic licence to me within the framework of naturalism. It's in both cases a conlang that's supposed to work like real langs.

While guides like the LCK are amazing (it's what got me into the whole thing) they can fail you in the following way: they let you forget that whatever steps you take are going to depend on the language you're intending to produce: if you're going for a language without any inflections then you can likely just skip the whole "nominal morphology, verbal tense, blablabla" bit of the grammar: you're not going to need it, for example. I like to peck and browse around rather than doing things very sequentially cause a lot of what interests me about conlangs is the way things come together (also cause I'm a bit of a nonlinear kinda dude): the classic example is if I have a lot of agreement in my nouns and adjectives, if every noun is inflected according to its role in the sentence then I probably can free up word order, right? except people don't work like that, often word order ends up being used to encode other things: maybe topicality, maybe putting the verb at the end is how you show deference, or imperativeness, or, what do I know, maybe number.

So a more minimalist sequence prescription, for a naturalistic artlang, might be something like

> get yourself a phonology: it doesn't have to be excruciatingly detailed yet, but it needs to establish clearly a) which are your phonemes, a.1) does the language encode information in other kind of phonological stuff? you know, tone, gemination, vowel lenght, creaky voice i dunno b) phonotactics. there's a bunch of ways to do this: for example you'd be better off explaining chinese in terms of "look: here's my allowed vowels, here's my allowed onsets, and here's my allowed codas. valid syllables are any combination of those". but other languages, like japanese, are better described as "look: instead of syllables let's talk about morae: a mora can be just a vowel, a consonant and a vowel, a consonant plus /j/ and then a vowel, the consonant /n/ on its own, or also the gemination of the whatever consonant came later: n and the gemination can't be the first mora of a word". some other language can be described as CV(ː)(T)(C), etcetera.
> fill up a bucket with pseudo-words: just, you know, allowed morphemes that feel to you like they sound like the language is supposed to sound: hint, if your phonotactics don't produce sequences of sounds that sound like the language you want to make, tweak the phonotactics: which phonemes are more frequent than others, which phonemes don't happen in a sequence, etcetera.
> figure out a few big picture things: how much inflection do you want on each part of speech and which categories are you gonna inflect for, if any: what's the overall word order, if you want one. what kind of morphosyntactic alignment are you going to want. do you want pro-drop? do you want any really funky stuff in your grammar, like nominal TAM? polypersonal agreement? maybe you want to go for polysynthesis? also consider dropping common features: maybe you have no pronouns? then what do fulfill the function pronouns perform? like, crucially, what do people answer questions like "who let the dogs out? when it is the asker that let the dogs out" some dummy verb? then you're probably going to want your verbs to somehow show personal information, like a verb "to be" that has a prefix that indicates second person singular or something. or you just have phrases or words like "he who asked the question!". in that case, those might as well be your pronouns, and you can slap a fancy name on that feature, maybe something like "periphrastic pronoun constructions" or something. phonological things can go here too! maybe you want some sort of tonal thing?
> try to translate a few basic phrases: you know, stuff like "hello, my name is William". don't word-for-word translate the phrases, either, but rather use them as opportunities to find things out which you don't know: take inspiration from natlangs you know! for example, chinese says this as "I am-called William (我叫william, I think, and yes, "the same" verb in one language can work the opposite in another language: in spanish we don't have 'to like' but rather 'to be liked by': me gusta means "it is liked by me", more or less.)
> make lexicon. this is its own thing and involves not just making words, but also derivational morphology, or words that mean different things depending on the part of speech it's behaving as (like, I don't like... know this but I like to think about things like the word "like" in english as an example)
> when you can translate a decent amount of stuff into the conlang, you can either a) go into more detail about like pragmatics, b) play around with sound changes to make a daughter lang, c) make a very detailed phonological account of how things are actually pronounced in the language
> add or remove features as you see fit
> get bored and start a new conlang.
User avatar
Pabappa
greek
greek
Posts: 577
Joined: 18 Nov 2017 02:41

Re: How to start making a conlang?

Post by Pabappa »

if this all seems a bit overwhelming, you can always just start small. you didnt say where your world is set up ... e.g. is this an alternate Earth or not?

even if not, if this is your first-ever language, you could start with something close to home, like a descendant of English or some other language you know well, even if this doesnt make sense on another planet.

then once you get comfortable with that, you can either start a new project or work your way through the first one such that it becomes independent of Earth after all.
Kavunupupis, šiŋuputata.
When I see you pointing at me, I know I'm in trouble. (Play)
User avatar
sangi39
moderator
moderator
Posts: 3024
Joined: 12 Aug 2010 01:53
Location: North Yorkshire, UK

Re: How to start making a conlang?

Post by sangi39 »

Pabappa wrote: 09 Jan 2021 13:21 if this all seems a bit overwhelming, you can always just start small. you didnt say where your world is set up ... e.g. is this an alternate Earth or not?

even if not, if this is your first-ever language, you could start with something close to home, like a descendant of English or some other language you know well, even if this doesnt make sense on another planet.

then once you get comfortable with that, you can either start a new project or work your way through the first one such that it becomes independent of Earth after all.
Definitely this! (alongside what Sal and Torco have said)

It definitely doesn't have to be perfect, or "exotic" or "exciting" or "new" first time round. You can have practice runs, start small and familiar, then as you read more, or develop different aims, either alter what you've got, or start over again with something else that incorporates that.

One of my favourite examples of this is Vrkhazhian by Ahzoh, which, from what I can remember, started out this vague notion of "wanting to make a language that employs triconsontal roots", and the first steps were sort of what you'd expect, i.e. that the vowels carried all the morphological work (this isn't how, say, the Semitic languages work, but they're often presented in a way that makes it look like that's how it works). Ahzoh, though, kept asking questions, kept researching and reading, would go backwards and forwards between bits here and bits there, and eventually Vrkhazhian really took form, and it feels "real" now. And keep in mind this took years, and probably won't end.
You can tell the same lie a thousand times,
But it never gets any more true,
So close your eyes once more and once more believe
That they all still believe in you.
Just one time.
Post Reply