Creating a language

[Aedilred]

All world-building leads back here eventually...

I have in mind to create a language, partly for world-building purposes but also for my own amusement. Has anyone given it a go? And does anyone have any useful pointers?

I've studied several languages to varying extents (albeit most of them Indo-European) and I think I have a reasonable layman's knowledge of how languages fit together and (to a lesser extent) evolve. I know the difference between a voice and a mood, and between a fricative and a plosive.

But my knowledge of linguistics is also pieced together from a combination of experience, a few "popular" books on the general or related subject, and internet browsing on Wikipedia or sites of similar depth. I suspect that in order to get anywhere with my project, I will at some stage, and probably earlier rather than later, need to know a bit more.

Does anyone have any recommendations for useful introductory texts to linguistics as an actual field of study? Or, even better, albeit somewhat optimistically, a good guide to creating languages that can talk me through some of the obvious-in-hindsight pitfalls?

[Aedilred]

Thanks for the link; I'll check it out.

These days I pick an existing language and use a variation of that. My elvish is a variant of Hindi, I've got a German-ish group, my islanders are speaking a simplified Irish Gaelic, and etc..

This is what I've always done before, but they're never quite right, and I've also virtually run out of languages with freely and easily accessible large lexica.

And yeah, I'm going into this with my eyes open about third-party appreciation. It's principally if not entirely for my own interest and satisfaction. The game to which the world-building relates may not even happen.

[Balyano]

On youtube Biblaridion, David Peterson, and Artifexian have some useful content. Also for a first language I suggest keeping it relatively simple, something you can pronounce and has grammar that doesn't stray too far from at least one of the languages you speak. (Though definately try out at one or two things not found in them!)

[Bohandas]

There's a computer program called DFLang that was originally designed for Dwarf Fortress modding which uses sample text from existing languages to generate made up words that sound superficially similar to the language used as the input. It even assigns the words meanings.

(On a side note, if you ignore the assigned meanings, its also a useful program for generating names, if you give it a bunch of names as the input instead)

http://dffd.bay12games.com/file.php?id=1738

[Aedilred]

Thanks for the help so far. I've acquired and read a couple of the books recommended by conlang.org, so I think I'm well on the way. I think I need to get a slightly better handle on phonetics before I get really stuck in: I'm probably about 75% of the way there, and nothing a few more sessions with Catford won't fix.

I think I will start with an experimental naming language which I don't intend to use seriously, to get the really basic mistakes out of the way, and then move on to the one I actually intend to get serious use out of. This will also give me a ready-made second language to use as a substrate/superstrate as necessary or borrow words from, as I try to evolve my primary one.

My current major hurdle is getting over the "kitchen sink" approach which is apparently common to early conlangers, where there is a tendency to throw in all the features they've heard of. As my primary will be a late-Neolithic language I want it to be relatively grammatically complex, so there is room to drop features (as is the norm). But I am reminded that although I have a pretty decent understanding of Latin grammar, that is not the same as actually being able to use the language with facility, because juggling dozens if not hundreds of endings in your head is challenging for those who are used to fairly un-inflected languages (like English). Which is likely to make actually using my language tricky if I stick to a similar level of complexity to Latin (which is in turn probably less inflected than PIE).

Equally, from a design perspective, a synthetic language is likely to be easier than an analytic one, if only because it cuts down to an extent on the semantic complexities. Especially if I don't want word order to be a critical feature, which I don't think I do. So I will need (for instance) a decent case system at minimum. And looking through the classical Latin case system I can only see one that I feel safe disposing of (vocative, and to the extent that it counts, the locative) if word order is going to be dispensed with. There's also one I'd like to add (essive). Which still leaves us with six, and probably even more endings to learn than for Latin. Yes, ancient Greek did away with the ablative too, but then I hate Greek.

Maybe I can cut down on the number of suffix declensions by using a wider set of noun-class prefixes (which I know I want to do anyway). Or possibly fold ablative and dative into the accusative and use particles instead - but that's getting into developments I would ideally want to come later.

I can see that this is going to be interesting.

[Aedilred]

Maybe start with a pidgin (Tok Pisin and Toki Pona are good references), and move it into a creole, and then expand it into a full-fledged language? If you use a script that evolves with the language you've got three for the price of one.

The problem with pidgins and creoles, from a creation perspective, is that they presuppose an existing language (or two). So I either have to start from a real-world language, which I want to avoid if possible, or create a "proper" language anyway.

Alternatively, I could start with a language which I treat as a pidgin/creole and develop the formative languages backwards from it, but I think that would be even harder than starting with the proto-language and evolving it.

[Vegan Squirrel]

You might find this tool helpful. It's a vocabulary generator based on the sets of syllables and phonemes you decide to use for your language. This one lets you input words from a language and run them through a bunch of sound/spelling shifts, which either lets you quickly build up dialects/periods of your language, or you can use it to take a real language and make it into something familiar but different. They take a little time to figure out, but I find them to be really neat tools, even though I never seem to get around to using them much.

Another helpful guide.

When I've toyed around with languages in the past, I get all excited creating the grammar, conjugations, and most basic words, then get stuck with the sheer amount of vocabulary you'd need to create to make it workable. So, the generators above help with that step, especially if you have a strong idea of the phonology of the language. Also, you can think of case endings, in particular, like prepositions in English, if that helps you wrap your head around the variety. We have a lot of prepositions, they just go before the nouns instead of attaching at the end. Obviously your language can have both (like German).

Anyway, I hope some of that is helpful, enjoy the journey! And if you're planning on using your own script for the language, it could prove useful to think about the transliteration to the Roman alphabet as you build it, since you'll probably end up using that for convenience sometimes.

[Aedilred]

You might find this tool helpful. It's a vocabulary generator based on the sets of syllables and phonemes you decide to use for your language. This one lets you input words from a language and run them through a bunch of sound/spelling shifts, which either lets you quickly build up dialects/periods of your language, or you can use it to take a real language and make it into something familiar but different. They take a little time to figure out, but I find them to be really neat tools, even though I never seem to get around to using them much.

Another helpful guide.

When I've toyed around with languages in the past, I get all excited creating the grammar, conjugations, and most basic words, then get stuck with the sheer amount of vocabulary you'd need to create to make it workable. So, the generators above help with that step, especially if you have a strong idea of the phonology of the language. Also, you can think of case endings, in particular, like prepositions in English, if that helps you wrap your head around the variety. We have a lot of prepositions, they just go before the nouns instead of attaching at the end. Obviously your language can have both (like German).

Anyway, I hope some of that is helpful, enjoy the journey! And if you're planning on using your own script for the language, it could prove useful to think about the transliteration to the Roman alphabet as you build it, since you'll probably end up using that for convenience sometimes.

Thanks for the links. I had already found and been using the vocab generator, as I picked up a couple of books and one of them is by the guy who runs the zompist site. It's been useful for my first language, where I am not too fussed about how it sounds, and am more interested in working out some of the noobish kinks in language creation before I get on with the one that I actually plan to use.

With regard to a script (again, for language #2), I have an idea of how I want it to look, but I need to work out my phonetics properly first. The simplest way to approach it would seem to be to create a 1:1 phoneme:letter alphabet and spell it completely phonetically. In which case I can also Romanise it relatively easily with sufficient diacritics and the like, so long as I keep it consistent.

You're not wrong about lexicon creation. I've been leaning on the vocab generator fairly heavily. I'm so far up to 200 words, although this includes quite a few derivations which I've developed rough schemes for (big, bigness, bigger, biggest, bigly etc.) This has however been enough for me to start translating the Babel text, and doing so has helped me with development of syntax as well as vocab:

Ke e irba hokuk pluda iait viserikit ke lait gaviskit.
Ke se ekmuoda ait vaiak tasoda zu e Gnohor, vaiak omfoda fravobos ho e dafnir Shinarot, e vaiak osdida hi.
Ke vaiak etarda iai po iazusa, ule po, skede asugis, ke klikefe aseb atrali. Ke vaiak orgamoda asugis piazin, ke orgamoda unges pohebon.
Ke vaiak etarda ia po iazusa, ule po, glude sulsos ke aludibos, pis abrot ikrodon e Ofsemarbot.

[Aedilred]

Alright, the Babel text is finished for the "practice language". This was a useful exercise, because it forced me to think about syntax and prompted creation of some new vocab. The total lexicon is still under 250 words, so far from a working language, but it was enough I think to expose a number of issues.

For instance I had been hoping to get away without a passive voice, but when confronted with it I couldn't think of another way to construct that sentence while preserving meaning. Of course, I could have just reworded it and said "this language doesn't have a way of expressing that" but at least I'm thinking about that.

I'm about ready to start on the principal language I think.

And the full Babel text for anyone who wants to try deciphering the language...

Ke e irba hokuk pluda iait viserikit ke iait gaviskit.
Ke se ekmuoda ait vaiak tasoda zu e Gnohor, vaiak omfoda fravobos ho e dafnir Shinarot, e vaiak osdida hi.
Ke vaiak etarda iai po iazusa, ule po, skede asugis, ke klikefe aseb atrali. Ke vaiak orgamoda asugis piazin, ke orgamoda unges pohebon.
Ke vaiak etarda ia po iazusa, ule po, glude sulsos ke aludibos, pis abrot ikrodon e Ofsemarbot.
Ke e Korsmok embadda tobos e paekis ke aludibos pik asagu asembot gluba.
Ke e Korsmok etarda, gahede, e ashihi plu iai, ke vaiak erb argamos iais viserikis, ke vaiak sagos a ubos se, ke via kasek pik vaiak krisnoda a ubos isibredon zu vaiakir.
Ule po, embade oi, ke hi urfnede vaiakit viserik, oet vaiak ke ka imarbos gaviskis ia-iazusit.
Sui, e Korsmok agenvoda vaiakis dako zu hi, az apmabor hokukit, ke vaiak erhoda a glu sulsos.
Sui, iot ul plu Babel, isu hi e Korsmok urfnoda viserikis hokukit: ke zu hi e Korsmok agenvoda vaiakis daku, az apmabor hukukit.

[MR_Anderson]

I have always found language, lettering systems, and syllabary & logographic communications interesting.

In the language you are making, are you using a 52 or 44 lettering system?

Are you creating any kind of letter to number coding system within the language?

[Aedilred]

I have always found language, lettering systems, and syllabary & logographic communications interesting.

In the language you are making, are you using a 52 or 44 lettering system?

Are you creating any kind of letter to number coding system within the language?

So far I have (iirc) 35 letters, because I've reduced the number of phonemes from English. and if starting from a 44-letter system I've only added one new letter: a distinction between clear and dark l (i.e. /l/ and /ɬ/. All of those cut so far are vowels or diphthongs. I am toying with culling /w/ and /dʒ/ too, and with adding alveolar trills.

I haven't yet created my own script as I want to get the bulk of the language nailed down first in something I can read easily. The limitations of the Roman script are however rapidly becoming apparent, particularly as regards vowels. I don't really want to add new symbols so I'm making do in transliterations with digraphs and diacritics, which is mostly ok (if a pain to type the diacritics) but I would like there to be a way to distinguish /ng/ from /ŋ/ visually without having to use the IPA character.

With letter to number coding, do you mean a straightforward A=1, B=2, etc? If so I haven't thought about it but there's no reason that it couldn't be done.

[Aedilred]

OK, having added a couple more consonants to be on the safe side, I have my list: 38 "letters" (or rather, phonemes) in total.

I realise at this point (to paraphrase Rosenthal) presenting these is a bit like providing a list of colours you intend to use in producing a painting, but still. I have started work on the lexicon, but barely enough to produce useful or interesting sentences, and the phonetics are important.

The "symbol" is the single-character version of the phoneme used when inputting data to word generators, and to assist me with confirming where something is two distinct phonemes next to each other, and where it's a digraph. They will also form the basis of the script when I come round to that. I have also provided the IPA to be on the safe side.

Note that the letter transliterated as "ɯ" has no equivalent in English that I'm aware of, and the trilled "r"s have no equivalent in standard English (although do appear in some accents and dialects), hence why they have a N/A in their column.

Romanised letter	Symbol	Phomene	IPA	As in English
ch	č	Affricate	tʃ	child
j	j	Affricate	dʒ	justice
l	l	Clear alveolar lateral approximant	l	light
ll	ł	Dark alveolar lateral approximant	ɫ	call
h	h	Glottal fricative	h	hussy
r	r	Post-alveolar rhotic approximant	r	wrong
n	n	Voiced alveolar nasal	n	never
m	m	Voiced bilabial nasal	m	mountain
b	b	Voiced bilabial stop	b	balance
z	z	Voiced coronal sibilant	z	zebra
dh	ð	Voiced dental fricative	ð	there
d	d	Voiced dental stop	d	dance
w	w	Voiced labio-velar approximant	w	woman
v	v	Voiced labiodental fricative	v	valiant
y	y	Voiced palatal approximant	j	yesterday
zh	ž	Voiced palato-alveolar fricative	ʒ	measure
rr	ᴚ	Voiced trilled fricative	r̝	N/A
ng	ŋ	Voiced velar nasal	ŋ	ring
g	g	Voiced velar stop	g	game
ɯ	m̥	Voiceless bilabial nasal	m̥	N/A
p	p	Voiceless bilabial stop	p	pottery
s	s	Voiceless coronal sibilant	s	style
th	θ	Voiceless dental fricative	θ	mythic
t	t	Voiceless dental stop	t	tang
f	f	Voiceless labiodental fricative	f	fine
sh	š	Voiceless palato-alveolar fricative	ʃ	shock
rh	ᴙ	Voiceless trilled fricative	r̝̊	N/A
c	c	Voiceless velar stop	k	cat

Vowels are a little easier.

Romanised letter	Symbol	IPA	As in English
ī	ī	i	feel
ū	ū	u:	food
ā	ā	ɔ	small
ē	ē	eɪ	day
ō	ō	əʊ	bone
i	i	ɪ	cricket
a	a	æ	cat
e	e	ɛ	bed
u	u	ʌ	fun
o	o	ɒ	dog

I'm not 100% sure whether I want to throw in the diphthongs /aɪ/ or /aʊ/, but I feel that ten vowels is already quite a lot. It's fewer than English, but English has unusually many. Plus I then run into the problem of how to write them, and the long vowels are already a pain to type.

[KragBrightscale]

Interesting stuff here, seems like you’ve made good progress.

Disclaimer: Although I’ve studied both Arabic and Mandarin immersively, I’m not that familiar with linguistics apart from personal experience.

Just wondering if you’ve taken into consideration the connections between language, thought, and culture, as well as the connection to the arts such as poetry.

If you are doing this as a hobby, you might find it interesting to read up on some of the discussions regarding the topics of linguistic relativism and linguistic determinism. For your homebrew this might help give your created language and culture another layer of depth/plausibility.

Are your nouns gendered? If so does that have an effect on the culture or gender roles? How does the language view time? Or describe locations and directions?

As for vocabulary, are you using a language as a base to determine which words you need? To illustrate: English has the words “goat” and “sheep”, while the mandarin word “Yang” (羊) can describe either (I think the one it’s refers to can be understood based on context). English’s words “cousin”, “aunt”, and “uncle” are not as precise as in Mandarin and Arabic, both of which have words indicating which side of the family, as well as other information such as who is older or whether blood related or not.

Does the language have a connection to the climate or context? Does it have more words to describe certain phenomena (example: precipitation words).


As you already seem to be well on your way to developing your framework, my suggestions may not be helpful at the moment. Maybe for your next one?

To create a language that feels more foreign (to English speakers at least), using a non-European language as a reference can have a lot of potential.

I’ve found Mandarin a really interesting Interesting language to learn. In particular the abundance of homophones, the non-phonetic writing system, and the grammar. Also: chinese characters! Lots of meaning packed into a small symbol!

Arabic, another language worth studying, has a really neat way of determining vocabulary. Most words can be traced to a 3-consonant root (past tense verb) and that is how dictionaries organize them. By inserting vowels, prefixes and suffixes (like following a formula) you can extrapolate and determine the rest. Which is great if you don’t want to come up with an exhaustive vocabulary list, just determine the roots and the rules.

Okay, got a little carried away...
Interested to see how your language develops and how/if you implement/use it in your homebrew.

Personally I like connecting magic to language...

[Aedilred]

I have actually thought about some of this a fair bit!

If you are doing this as a hobby, you might find it interesting to read up on some of the discussions regarding the topics of linguistic relativism and linguistic determinism. For your homebrew this might help give your created language and culture another layer of depth/plausibility.

Are your nouns gendered? If so does that have an effect on the culture or gender roles? How does the language view time? Or describe locations and directions?

Re: gender, I have been partially persuaded that for all the superficial attractiveness of the suggestion that language gender reflects cultural gender roles, it doesn't bear up under serious scrutiny, given how many extremely patriarchal societies use genderless language (or even, I think, the occasional feminine-preference one). Nevertheless I do know that gendered pronouns are a problem for a lot of people and I see no reason not to resolve that since I'm starting from scratch. So there are four genders: masculine, feminine, neuter-animate and neuter-inanimate, which allows for non-binary singular pronouns that are nevertheless personal (i.e. neither the occasionally clumsy "they" nor the uncomfortable "it").

Time and directions are something I've been giving some thought to but haven't settled on. As regards compass directions, I'm not sure whether to pick my "primary" and then determine the others by reference to it (i.e. something like East, Left-East, Right-East, Opposite-East) or just have a unique word for each.

I haven't really touched time yet because I want to get a handle on the setting, number of moons, seasons, days in a year and so on before it makes sense to start dividing it up into weeks, months, years, etc. I did toy with making aspect more of a feature, but I'm more comfortable thinking in terms of tenses for now: aspect still gives me a bit of a headache.

As for vocabulary, are you using a language as a base to determine which words you need? To illustrate: English has the words “goat” and “sheep”, while the mandarin word “Yang” (羊) can describe either (I think the one itÂ’s refers to can be understood based on context). EnglishÂ’s words “cousin”, “aunt”, and “uncle” are not as precise as in Mandarin and Arabic, both of which have words indicating which side of the family, as well as other information such as who is older or whether blood related or not.

Does the language have a connection to the climate or context? Does it have more words to describe certain phenomena (example: precipitation words).

One of the things I've already done is with connection to the local environment and in particular trees. The words for "oak", "ash", "elm", "birch" etc. all go back to IE and have been used to try to identify the IE urheimat. But I want my people to start out from a rainforest environment, so the root trees are kapok, ebony, teak, walnut etc. And instead of (as in older varieties of English) "meat" being synonymous with "food", it uses the root for fruit instead. I will definitely be applying similar principles to other objects, animals, etc.

With family I did consider doing something radical, or more detailed as in Mandarin, but I eventually concluded that given this is an early Bronze Age, pre-literate society where communities are quite small, kinship groups will both have a fair amount of overlap and be neither particularly relevant nor well-known beyond living memory, so I ended up limiting specifics to the nuclear family and having more general words for "relative" beyond that.

To create a language that feels more foreign (to English speakers at least), using a non-European language as a reference can have a lot of potential.

IÂ’ve found Mandarin a really interesting Interesting language to learn. In particular the abundance of homophones, the non-phonetic writing system, and the grammar. Also: chinese characters! Lots of meaning packed into a small symbol!

Arabic, another language worth studying, has a really neat way of determining vocabulary. Most words can be traced to a 3-consonant root (past tense verb) and that is how dictionaries organize them. By inserting vowels, prefixes and suffixes (like following a formula) you can extrapolate and determine the rest. Which is great if you donÂ’t want to come up with an exhaustive vocabulary list, just determine the roots and the rules.

Definitely! My forays into non-Indo-European languages have generally ended with frustration. Hungarian I just couldn't get on with at all and for both Arabic and Hebrew I have so far found the script an insuperable barrier. But I have tried to look at ideas from non-IE languages all the same. So while it will almost certainly end up being more similar overall to IE than to anything else, I think (or at least hope) there are some completely non-IE features.

One of the things I am incorporating is a noun class system which is essentially borrowed from Setswana. Indeed my discovery of that was one of the things that prompted me to create a whole language. By varying prefixes on nouns (and then by extending the same principle to adjectives, and eventually verbal mood) a whole range of related concepts can be developed from the same root. This has allowed me to "unpack" a lot of things from a single English concept, and gives me room to expand from certain roots later. For instance, the idea of "home" can, by varying the prefix, express the building you live in, people in your household, the geographic district you live in, the feeling of being "at home", the quality of homeliness, etc.

The downside of this is that even simple words are turning out to be quite long. Starting from a single-syllable root, by the time a case marker and prefix are added, three to four syllables is the norm. That would definitely make poetry an interesting challenge.

A couple of the other things I'm trying to do, largely to add to the feeling of difference, are (firstly) to remove phrases with a verbal copula, like "he is clever" and instead verbalise adjectives so there'll be a verb for "is clever" derived from the adjective;, and (secondly) to remove adverbs, instead expressing the same concepts with nouns. e.g. "She thought with quickness" instead of "she thought quickly". As and when I "evolve the language that might be something I look at, to see whether it's working and whether speakers would start creating shortcuts.

[aimlessPolymath]

One trick you could borrow from Hebrew (I don't know much about other languages or about conlangs) would be to have each root word be consonant-based (hebrew has three-letter roots); this gives you space to vary the vowels in addition to adding prefixes/suffixes, hopefully reducing the length of words.

[Aedilred]

One trick you could borrow from Hebrew (I don't know much about other languages or about conlangs) would be to have each root word be consonant-based (hebrew has three-letter roots); this gives you space to vary the vowels in addition to adding prefixes/suffixes, hopefully reducing the length of words.

This was something I thought about, initially with a view to incorporating some kind of ablaut system, but I decided that regularising it in a satisfactory manner would be a bit too much work.

I don't think it's wholly unwieldy and in fact even the longer words don't tend to exceed four syllables, but they do look a bit long.

All, an ivull fariyell ācug sūwegog all ācug sūwepāg ēvath.
All urel baning an faricdīd eng tirrtovan, faraduzhūng farūd shinarīg el ādhūyon, all baning ālt dundāyon.
All baning banūdcārēng epāllayon, ladorgyūng ichami, all yūng vemenganūd ilbōthi. All ladorgyūng lacimūllch nēmeyon, all lepērhūll lemallpūllch nēmeyon.