The most difficult things about conlanging

[linguistcat]

What do you find the most difficult about conlanging?

Personally, it's keeping track of things in a way that makes sense later or that I don't want to change. It's hard for me to keep things organized but I also don't get far on any projects if I don't stay organized.As a subset of this, I'm trying to decide how to mark tone changes in words in a way that makes sense to me (regardless of what "official" standards there are) since my conlang is going to have a modification of a pitch accent system.

[malloc]

Creating vocabulary items, simply because any language (apart from oligosynthetic ones) needs thousands of them, and they must be reasonably distinct yet feel like they belong in the same language. The biggest roadblock to conlanging for me is getting around to creating the actual vocabulary after I have worked out the phonology and morphology and such.

[evmdbm]

If you do words as you need them or just a few at a time it must be easier.

I think it'll be the syntax and working out idiomatic constructions that aren't just copies of English or French or whatever

[alice]

Trying to come up with features which (1) are interesting, (2) are unusual, and (3) have not yet appeared in one of Zompist's conlangs.

[masako]

Creating vocabulary items, simply because any language (apart from oligosynthetic ones) needs thousands of them, and they must be reasonably distinct yet feel like they belong in the same language.

Agreed. 100%

Trying to come up with features which (1) are interesting, (2) are unusual, and (3) have not yet appeared in one of Zompist's conlangs.

I don't want this to sound insulting, but 2 and 3 are stupid.

Unusual and interesting are not directly linked. One of the most ubiquitous things in Indo-European langs is the verbal conjugations, and I find them fascinating. Also, given that every known writing system evolved from some form of pictographic ancestor, a script that uses ideo/pictograms isn't very original, but still captivates me. On the flip side, Pirahã is unique in that it reportedly only has lemma for the numbers "1" and "2" which are phonologically similar...honestly, that bores the shit out of me. Show me an indigenous language that has complex mathematical descriptions for landscape features, like Navajo...now that shit is interesting.

#3 is stupid just because Zomp has quite a compendium...it would be an insane challenge for anyone to avoid all of those features.

[Zaarin]

For me it's syntax. I love building vocabulary--it's fun, honestly--and I can even manage with morphology, but then I sit down to use the conlang and find I have no clue how sentences are supposed to work in the language.

[chris_notts]

Depends what you care about, I guess, and how deep down the rabbit hole you want to go.

My problem is that everything is related to everything else. Thinking up a bundle of features you think are cool is easy, but then they all need to work together and form a realistic looking system. A few natlangs are kitchen sinks, but most aren't, so you want a system of complementary morphemes and structures. Oh, and the lexicon itself contains patterns, which differ between languages but are much less studied than syntax and morphology. And it's so easy to just forget or copy how things work in your native language(s), so to get everything right you need to be an expert in everything. And of course there are lots of strong but not exceptionless typological generalisations, so if you want something that feels real not just possible you need to have a realistic mix of the exceptional and the common.

For example. Maybe you've decided to have a nice symmetrical voice system (= multiple transitive voices), like a direct-inverse system, or a Tagalog style trigger system, or whatever. Fine. But now you've got to decide what that means for every construction in the language. Gapping and coordination, relativisation, focus and clefting, topicalisation, subordinate and complement clauses. What are the pivots and coreferential restrictions in all of these constructions? You might think the answer is easy, but it's not: in Phillipine languages and for that matter in Algonquian languages some constructions are sensitive to the A/S, some to the privileged argument, and some to neither.

Similar for applicatives - do the applicative objects gain all properties of the base object? Does the base object lose all its properties? And if you have multiple constructions, do they all pattern the same or differently? Is promotion required by certain constructions?

And it's the same for everything. It's a bit like writing a novel - everything interacts with everything else, and there's a continuous revision involved to try to create an organic whole.

[masako]

It's a bit like writing a novel - everything interacts with everything else, and there's a continuous revision involved to try to create an organic whole.

[akam chinjir]

My problem is that everything is related to everything else.

Yeah, this, together with the fact that everything's the tip of a diachronic iceberg (and it's icebergs all the way down, practically speaking).

[chris_notts]

My problem is that everything is related to everything else.

Yeah, this, together with the fact that everything's the tip of a diachronic iceberg (and it's icebergs all the way down, practically speaking).

For that reason, I normally put a lot less effort into the diachronics, since it's a lot of work even getting a snapshot right. For me, a really good conlang grammar should be comparable in detail to works like these, which each run to hundreds of pages:

https://www.cambridge.org/core/books/gr ... ABD87C456C
https://www.aikhenvaldlinguistics.com/manambu-language
https://www.amazon.co.uk/Grammar-Semela ... 052114499X
https://www.amazon.co.uk/Jarawara-Langu ... 0199600694

...and that's enough work without also documenting a proto-language. What I usually do is just add suggestive hints to the synchronic shapes of morphemes and allomorphs, e.g. alternations which reflect common sound changes cross-linguistically, without fully working out the historical changes.

[akam chinjir]

For that reason, I normally put a lot less effort into the diachronics, since it's a lot of work even getting a snapshot right.

One of my stumbling blocks has actually been that I'll start thinking I need an ancestor language, but then get caught up in the details and complexities of that so much that it starts wanting its own ancestor. Heck, Akiatu was first conceived as a protolanguage so simple in its phonology and morphology that I wouldn't get caught up in trying to derive it from something still earlier. (But of course now I'm trying to retrofit an ancestor onto it because of some deep dissatisfactions, with lexical stuff especially. And meanwhile the language with which the project started is thousands of years in the future, and not genetically related.)

I've actually found the forum format really helpful in dealing with the interconnectedness of all things, since each post can be short and provisional. Among other things, it's a pretty good way to throw out ideas, and then see which of them end up playing well with later ideas, and which of them sort of fall by the wayside.

For me, a really good conlang grammar should be comparable in detail to works like these, which each run to hundreds of pages:

https://www.cambridge.org/core/books/gr ... ABD87C456C
https://www.aikhenvaldlinguistics.com/manambu-language
https://www.amazon.co.uk/Grammar-Semela ... 052114499X
https://www.amazon.co.uk/Jarawara-Langu ... 0199600694

...and that's enough work without also documenting a proto-language.

Yeah, though I think there's also a sweet spot around 50-60 pages---long enough that it can go into some depth, short enough that reading it with some care isn't a massive undertaking. (A lot of the grammars in the Languages of the World series are about that length, and I love it.)

[chris_notts]

Yeah, though I think there's also a sweet spot around 50-60 pages---long enough that it can go into some depth, short enough that reading it with some care isn't a massive undertaking. (A lot of the grammars in the Languages of the World series are about that length, and I love it.)

I agree, a good survey is also really good. One issue I've had though is the tendency of that kind of mini-grammar to focus too much on simple, single word surface phenomena (phonology, morphology) at the expense of syntax. I have quite a few regional or language family surveys books, and while they do cover those particular areas pretty well they describe much less well, or at least more variably, things like relative clauses (unless marked by morphology), subordinate clauses, coordination, control, raising, scope effects... etc. But that's a problem for longer grammars as well. I have some longer grammars of Australian languages on my book shelf that barely mention syntax at all. To be fair, that's partly because those languages are very morphology heavy with very free word order, but that doesn't mean they have no syntax whatsoever.

This is also a common problem with conlang grammars: they ignore most things that aren't phonology or morphology.

[Frislander]

One of the hardest things for me is trying to be as enthusiastic about all areas of my conlang, like being as interested/having as many ideas regarding the syntax as the phonology. This has often been a problem for me in the past: I come up with this great phonology, start a conlang... and then realise I have no idea what I want the morphology to do, and end up stuck with nothing and scrapping it.

Trying to come up with features which (1) are interesting, (2) are unusual, and (3) have not yet appeared in one of Zompist's conlangs.

Like masako I agree that (3) is silly, but for a different reason. I don't recall seeing a conlang of his featuring full-blown polysynthesis, fewer than 5 vowels, ejectives, or indeed only 1 stop series, CV syllables, etc. Which is fine I don't expect even the zomp to have created fleshed out versions of every conceivable kind of conlang under the sun, but I think it's also quite clear that the main focus of his work has been on one branch the Eastern family, and everything else is essentially rotating around that centre of mass. And a lot of the non-Eastern stuff is decidedly "weird" in very specific ways even by the standards of the world's languages as a whole, like Munkhâshi definitely does not feel at all natural to me (like I know the ktuvoks and their stuff allows for some creative liberties, but still), and I'm not too sure about Kebreni or Old Skourene on that point either. There's plenty of stuff that's found in natural languages which I haven't seen yet in there.

[Neon Fox]

Elkaril has ejectives. But the greater point stands, certainly.

Oh no, I'm wrong, it has implosives. Ah well.

[xxx]

the difficulty is the same as for a marathon...
keep your motivation to sustain the distance...
fortunately the pleasure of discovery is a powerful engine...

[Vardelm]

Thinking up a bundle of features you think are cool is easy, but then they all need to work together and form a realistic looking system.

This.

I also find it really difficult to create affixes for inflection that are aesthetically pleasing. I need to know more about diachronic phonology as well so that I can make those affixes less regular but still plausible.

For some reason, I find it fairly easy to come up with words. I find real languages that are similar-ish, look up related words in those languages, and then sort of mash them together in a way that fits my conlang's phonology. Works well for me so far.

[TurkeySloth]

Good grief, Charlie Brown!

By far, the most difficult thing for me was finalizing the phoneme inventory. I think I went through, at least, seven iterations before settling on my now-final version.

[Vardelm]

Good grief, Charlie Brown!

By far, the most difficult thing for me was finalizing the phoneme inventory. I think I went through, at least, seven iterations before settling on my now-final version.

Of all the conlang grammars I've read in my lifetime, your's is the CharlieBrowniest.

[Pabappa]

For Poswa, syntax is the most difficult thing. Word generation is easy because I prioritize Poswa above all other languages, and therefore if a Poswa word conflicts with something in another language, Poswa wins and I have to remake the other word instead. But syntax is difficult because it's so unlike IE. I downloaded an Inuit grammar a while back since its the language thats grammatically closest to Poswa, but I prefer to work blindly and therefore Ive only looked at it once so far when I was really stuck on some minor thing to do with word order and wanted to see what my options were.

For the other languages, vocabulary is difficult because I have to check every word to make sure its etymology doesnt conflict with something I need for Poswa. That and I find it really difficult to work on a language that doesnt have a phonology I love, and I can only love one type of phonology. I actually had a language with >1000 words called Khulls which is effectively dead now because Ive scrapped so much of the vocabulary it depends on and may not even be able to keep the same phonology.

All of my planet's languages fit into a single sprachbund, comparable in size and diversity to Europe, Australia, or SE Asia. Thus, the grammar of one language is always fairly similar to all the others, and even though not all of the languages are polysynthetic, I think that once Ive found a solution for any given construction in Poswa, I can extrapolate it to the entire planet, since in most cases, its basically a 1-dimensional scale with Poswa at one end and IE-like languages at the other. Thats why Im busy right now coming up with things like grammaticalized obeidence that I dont think Poswa will use very much .... just teh fact that it exists helps me with the other languasges.

[zompist]

I don't recall seeing a conlang of his featuring full-blown polysynthesis, fewer than 5 vowels, ejectives, or indeed only 1 stop series, CV syllables, etc.

FWIW, most of these features exist, but not in worked-out languages: the Itsenic languages are polysynthetic; K'aitani has ejectives; Luxajia is CV (Wede:i comes close). Tžuro was intended to be three-vowel but didn't quite make it. If you looked at the major languages of Earth you wouldn't get all possible features either...

As for the most difficult bit of conlanging— for me it's the lexicon, just because it takes so long. It's best when it expands as I work out the syntax and translate sample texts, but there always remains a slog of word creation at the end. I don't hate it or anything, but it's hard to do for hours on end.