General problems in language design

[Ναθια]

You most certainly do have to know the creator's goals—otherwise you run the risk of saying things like "This poem is awful, it doesn't even have any airbags!" or "This car is terrible, its rhyme scheme is all wrong."

That you have a point - yes, you do have a point - doesn't invalidate the notion that general design principles apply across categories. Conlangs are not as different as poems and cars, and there are likely to be some strategies or combinations which should work better than others whether in an auxlang or artlang.

Moreover, even poems and cars should both be complete, and demonstrate some symmetry as well as efficiency. And of course their elements should work harmoniously with one another. In a car, having elements work harmoniously means the batteries should carry enough charge for the starter, and an engine large enough for the overall weight of the vehicle. In a poem, having elements work harmoniously generally means selecting words that fit a rhyming scheme or meter, and avoid word pairs with awkward or repetitive clusters; a master poet could also determine which kinds of words would best fit the poem's meter, and even find ways of fitting difficult words into that poem as careful flourishes.

But what does "having elements working harmoniously" mean in the context of a constructed language? I don't know. Can someone tell me how to achieve it?

[Micamo]

But what does "having elements working harmoniously" mean in the context of a constructed language? I don't know. Can someone tell me how to achieve it?

This, unfortunately, is a little hard to explain. It's something that more has to be felt than calculated.

Where I can start, though, is that you should avoid languages like Ithkuil. Ithkuil is the result of the author taking as many morphological distinctions as he could think of and cramming them all into a single language. Like a soup that includes everything in the kitchen, it's a miracle it's even edible.

What you want is for the little details (the micro-features) to either all come together to form broad patterns (macro-features), or the reverse: For a bunch of seemingly unrelated little details to logically follow from a single underlying pattern that simply has non-obvious consequences. Natlangs are a wonderful source of inspiration for this: Tlingit, for example, has a set of prefixes which seem to irregularly undergo phonological reduction. But when you look deeper, you'll find that the rule is actually that every prefix within the inner conjunct domain that can undergoes this reduction, and the ones that can't are prevented from doing so by the normal phonotactic rules that govern the rest of the word. This is harmony.

I (and many other conlangers) absolutely delight in finding these sorts of patterns. They're what make natlangs so beautiful, and we try to imitate this beauty in our own languages.

[Trailsend]

Conlangs are not as different as poems and cars

You'd be surprised

Moreover, even poems and cars should both be complete, and demonstrate some symmetry as well as efficiency. And of course their elements should work harmoniously with one another. In a car, having elements work harmoniously means the batteries should carry enough charge for the starter, and an engine large enough for the overall weight of the vehicle. In a poem, having elements work harmoniously generally means selecting words that fit a rhyming scheme or meter, and avoid word pairs with awkward or repetitive clusters; a master poet could also determine which kinds of words would best fit the poem's meter, and even find ways of fitting difficult words into that poem as careful flourishes.

But a car built to be part of a museum exhibit need not have a battery at all, and a free verse poem is explicitly not supposed to have rhyme and meter. Similarly, I think you will find that for any criterion you propose, a conlang can be created that does a very good job of violating it.

But what does "having elements working harmoniously" mean in the context of a constructed language? I don't know. Can someone tell me how to achieve it?

It depends on the purpose of the conlang—what makes the elements of many auxlangs and engelangs fit together harmoniously would often be considered a flaw in an artlang (e.g., self segregating morphology, extreme regularity, etc.). If you, with Micamo, are part of the large body of conlangers that enjoy finding hidden patterns, then the kind of behavior she describes in Tlingit could be a sign of harmoniousness. But if your language is intended to have very clear and obvious grammatical systems that are apparent at face value, the same behavior could be a flaw.

For your language specifically, some criteria that might be useful are:

1. A well-distributed phonological system, to aid with comprehension in noisy environments
2. A morphosyntactic system that provides creative (but not unwieldy) solutions to your family's communicative needs
3. A semantic system that specifically addresses areas of communicative value to your family
4. A general aesthetic that is appealing to your family members

It sounds like you have figured out (1) on your own—you want to avoid having words that are likely to occur together or in the same semantic field sound too similar to each other, having too little phonetic variation within and across words, etc. (The exact thresholds of "too similar" and "too little" are difficult to specify, which is probably why you've had a hard time getting advice on whether to expand your set of phonemes; a quick glance at the phoneme inventories and allowable syllable structures of Japanese and English would suggest that Japanese has substantially less phonetic variation than English does, but this is obviously not a problem for Japanese speakers, perhaps due to considerations that an English speaker might not think to look for at first.)

(2) is where many conlangers have the majority of their fun. The criterion to not be unwieldy is where Micamo's comment about Ithkuil applies; if you try to apply every bell and whistle you can think of to your language's morphology and syntax, you risk it becoming unfeasible for day-to-day use, bogged down with superfluous features. To return to the channel metaphor, this would mean digging extremely deep morphology and syntax channels, rather than distributing the informational load across semantics and pragmatics as well. (Information conveyed via the pragmatics channel is often the source of ambiguity, polysemy, and other phenomena that Ithkuil was explicitly designed to avoid, which is why that language had such fantastically deep morphosyntactic channels.) I would suggest taking a highly practical approach—if you find yourself frequently needing to make a particular distinction, then encode that distinction inflectionally or syntactically. (This can be a little tricky, since you first have to become aware that you are making the distinction, and if it is buried in English's pragmatic channel, you might not be conscious of it without some careful introspection.) For distinctions that are useful but not quite so common, a semantic or derivational solution might be better.

(3) has to do with your language's criterion that it be unique to your family. There are things your family talks about that are unique to your family, so a custom-fit language should be tailored to address them. I don't know what those things are, but there's lots of room for some fun creativity here.

I can't be much help with (4). I happen to be a fan of "l" and "th" sounds, but I know some other members of the board don't like them. (These people are crazy, but we humor them for politeness's sake.)

[Visinoid]

Ναθια:

What's your language background? (Spoken languages at work, at school, at home)(Learned languages)(Foreign family)

[JohnQPublik]

...Ithkuil is the result of the author taking as many morphological distinctions as he could think of and cramming them all into a single language. Like a soup that includes everything in the kitchen, it's a miracle it's even edible.

In the eight years since I first published Ithkuil on the web, I have never had any problem whatsoever with anyone offering a subjective personal opinion about my work, such as Mr. Micamo's comments about "soup" or my work's "edibility." However, I must now, for the first time, draw the line when someone mis-represents me personally as to the motive or purpose of my work.

I have not spent thirty years of painstaking effort to produce a language designed to more deeply reflect human cognition in a morpho-phonologically concise manner, only to have someone so casually, presumptuously, and discourteously imply my work "is the result of the author taking as many morphological distinctions as he could think of and cramming them all into a single language."

I invite Mr. Micamo and anyone else who may have a similar shallow, superficial, and erroneous understanding of my work to take the time to read the Introduction to the Ithkuil website in order to see, in fact, what Ithkuil really IS the result of.

--John Quijada

[Micamo]

I've read your entire presentation on Ithkuil. Multiple times, in fact. It has many creative ideas for morpho-semantic distinctions and is thus a great place to get inspiration. However, result is not the same thing as intent. I don't believe Ithkuil to be useful for anything other than that.

My first problem is, like Ladaan, it's an attempt to address a problem that doesn't actually exist, at least not to the extent you make it seem. Lemme quote your introduction so I can show you what I mean (My argument more or less generalizes to all obligatory inflectional categories in the language):

For example, in sentence (a) we have no idea whether the boy chose to roll himself down the hill or whether he was pushed against his will. (In formal linguistic terms we would say it is unknown whether the semantic role of the subject ‘boy’ is as agent or patient.) And yet knowing which scenario is correct is crucial to understanding the speaker’s intent in describing the action.

There's a third scenario here you neglect to mention: That the speaker does not know, nor care, whether the boy intentionally rolled down the hill or not. There's the possibility that within the current discourse this particular piece of information is simply irrelevant, and it's a waste of complexity to include it.

Furthermore, in the case that the boy's semantic role in the event IS relevant, there are easy methods available to speakers to include this information:

"The boy accidentally rolled down the hill."
"The boy decided to roll down the hill."
"The boy was pushed, and then he rolled down the hill."

The real problem to which you're referring, is that because the information is optional and not obligatory, it's possible for a speaker to leave out such information by mistake even though it is relevant to the current discourse, and that warrants the extra cost of complexity in forming sentences to make such information obligatorially marked.

This obviously can only be so true, as it applies to every conceivable piece of information about the event being described that could possibly be relevant within the context of a larger conversation: Whether the boy was wearing shoes or not, what color his hair is, the last time he clipped his toenails, what he wants to be when he grows up... Why should semantic role be obligatory information to express in a sentence, but not every other potentially relevant pieces? Why is the "line" of obligatorially expressed categories in Ithkuil drawn where it is, rather than anywhere else?

[Visinoid]

For example, in sentence (a) we have no idea whether the boy chose to roll himself down the hill or whether he was pushed against his will. (In formal linguistic terms we would say it is unknown whether the semantic role of the subject ‘boy’ is as agent or patient.) And yet knowing which scenario is correct is crucial to understanding the speaker’s intent in describing the action.

There's a third scenario here you neglect to mention: That the speaker does not know.

The real problem to which you're referring, is that because the information is optional and not obligatory, it's possible for a speaker to leave out such information by mistake even though it is relevant to the current discourse.

I'm curious about that topic. :3 Now, I'd ask to the author of Ithkuil if he could translate "The boy rolled down the hill" without the speaker knowing whether the boy is the agent or patient.

[Khemehekis]

2. If you don't need ten pages to explain how your case system works, then your language is too simple for my liking.

:Goes off to count the number of pages on the case system in my Kankonian grammar:

Wait, what case system?

[Curlyjimsam]

You most certainly do have to know the creator's goals—otherwise you run the risk of saying things like "This poem is awful, it doesn't even have any airbags!" or "This car is terrible, its rhyme scheme is all wrong."

But do you know the goals in these cases or do you merely infer them, on a very broad basis, from the work itself and the sociocultural context? The latter, I think - and this can apply to conlangs too, at least in theory. Perhaps there is a problem, though, in that there's maybe less agreement on what conlangs are for than there is for cars or poems.

[Systemzwang]

You most certainly do have to know the creator's goals—otherwise you run the risk of saying things like "This poem is awful, it doesn't even have any airbags!" or "This car is terrible, its rhyme scheme is all wrong."

But do you know the goals in these cases or do you merely infer them, on a very broad basis, from the work itself and the sociocultural context? The latter, I think - and this can apply to conlangs too, at least in theory. Perhaps there is a problem, though, in that there's maybe less agreement on what conlangs are for than there is for cars or poems.

To some extent there's information available in where in the classification scheme of auxlang||engelang||artlang||a posteriori lang||... etc it goes, but that is naturally rather open-ended information.

Some conlangers do provide a fairly good explanation as to their intentions in some kind of Introduction or Foreword to the grammar. I think there's few good enough conlangers to indicate it solely by the work itself, but I imagine there's at least a dozen of them, though.

[Zumir]

Really, a lot of the guidelines that apply to artlangs don't apply to engelangs. Unless one of the creator's main intentions with a language was to have it be aesthetically appealing, attempting to apply them is a bit like criticizing a non-fiction book for not having enough character development.
Although, being able to tolerate ambiguity is essential no matter what. If Ithkuil can't, then I'm with Micamo 100 percent.