Hallo conlangers!

On 01.06.2015 20:28, Jeffrey Brown wrote:

> Recently, on the Conlang forum, there was a thread that mentioned, in the
> midst of a different discussion, the question of whether or not the grammar
> of a conlang was naturalistic. This is a question I have pondered for some
> time, as I am much more fascinated with the syntax and morphology of a
> language, than with its phonology or its conculture.
> When conlangers speak of naturalistic conlangs, it seems to me that they
> speak primarily of the way the language sounds, and of whether the
> phonology has the “feel” of having developed naturally over a period of
> time, because of the judicious application by its creator of fictional
> diachronic sound changes.
> I think that conlangs grammars can also have a naturalistic “feel” as
> opposed to an artificial one. I cannot define what makes this so, other
> than it is a personal, subjective, aesthetic sense of the language’s
> structure, which I can perceive if I attempt to use the language for speech
> or translation. So, what follows is purely my own personal opinions of the
> naturalism of the grammars of various conlangs.
> I. Relexes
> Although everyone disparages relexes, one thing about them is that their
> grammars are naturalistic. After all, it is just a relexification of a
> natlang, and a natlang, by definition, has a natural grammar.

Depends on *how* you relex it.  A badly-done relex can create unnatural 
patterns from natural ones.  The source language may have an alternation 
between two allomorphs that seem arbitrary but actually have a common 
source (example: the Ancient Greek accusative singular ending, which is 
-n after vowels and -a after consonants); the relex may turn this into 
something that is just plain arbitrary and unnatural.  But overall, 
relexes of natlangs are likely to be quite naturalistic, at least at 
first glance, and definitely more so than ...

> II. Mathematical Grammars and Engelangs
> Any grammar based on predicate logic, or computer science, or any branch of
> mathematics, does not have a naturalistic feel. Examples are Fith, Loglan,
> Lojban, and Ithkuil (sorry JQ). This does not mean that these conlangs are
> inconsequential or unworthy in any way, merely that when I try to say
> something in that language, it doesn’t feel like I’m speaking a language,
> it’s rather like I’m trying to solve an algebra problem in my head. I am
> aware that are a some people who can speak Lojban more or less fluidly, but
> there are also some people who can quickly solve algebra problems in their
> heads and that doesn’t make algebra a language.
> I suspect that my perception is based on the fact (or supposition) that the
> parts of the brain that handle language are different than the parts that
> handle mathematics.
> You will note that these are all engelangs. In fact, I cannot think of a
> single engelang that has a naturalistic grammar. But that’s not the point
> of an engelang, anyway, is it?

Yep!  Not to disparage engelangs, but they are usually far removed from 
natlangs.  I have never seriously tried to use any, so I cannot confirm 
nor deny what you say about what it feels using them.  But I can imagine 
that using an engelang feels more like doing math or computer 
programming than like conversing in a natural language. (At least, as a 
former computer scientist, I have done quite a lot of programming, which 
involves using programming languages, of course, and it *never* felt 
like talking to the machine as if it was a human being.  It rather felt 
like building something from LEGO, or conducting a chemical 
experiment.)  It is at any rate true that mathematics and language are 
different skills using different circuits in the brain.

> III. Auxlangs
> The most common auxlangs are based on European languages. Their grammars
> are really a regularization and simplification of a Romance grammar, or at
> least of an IE grammar. I find these extremely easy. I can read texts in
> auxlangs I have never studied. And I find their grammars to have a very
> naturalistic feel.

At least, the mainstream IALs, such as Esperanto or Novial, feel much 
more naturalistic than engelangs.  They are more regular than most 
natlangs (though there are some highly regular natlangs around), but yet 
they work basically the same way as the natlangs they are based on.  
Nevertheless, these languages feel more like simplified models of "real" 
languages than like "real" languages.

> Now, one could retort, that because English is my native language, or
> because I speak several European languages, Esperanto and the like feel
> naturalistic to me because their grammars are similar to English or
> Spanish.


(It may be noted that the term "naturalistic" as used by auxlangers is 
to be distinguished from the homonymous term used by artlangers which 
this discussion is about.  Basically, "naturalistic" means that not just 
roots and affixes are borrowed from natlangs as in Esperanto, but whole 
words.  But that does not really matter here.)

> But, I have also learned Arabic and Hawaiian, and the grammars of
> those languages are decidedly not IE in structure, yet their grammars feel
> natural to me. I have also studied Japanese, Chinese, and Blackfoot (though
> I don’t claim more than a smattering of knowledge of them), and the
> grammars of those languages, too, feel natural.

They do!

> There is something about natural language grammars, their combination of
> features or their structure, which, in my opinion, “kicks” the processing
> of the language into Broca’s or Wernicke’s area (or wherever), rather than
> the part of the brain that handles general symbolic manipulation.

Seems like that.  Yet, it is hard to put the finger on it.

> IV. Minimalist Languages
> I love Toki Pona. Thank you, SL! But it doesn’t have a naturalistic
> grammar. It’s a toy grammar.

Yes.  Toki Pona doesn't feel naturalistic.  It lacks the richness of 

> V. Exolangs, Written-only langs, or any Non-sequential lang
> Human language is oral/aural and is produced by a single channel (one
> larynx, one channel). Any conlang that doesn’t conform to those
> restrictions ends up with a non-naturalistic grammar, in my opinion. But,
> maybe that’s part of the point of an exolang, to come up with an “inhuman”
> grammar, isn’t it?

Agreed.  Yet, there are conlangers who use "They are not human!!!" as an 
excuse for a poorly-designed conlang.  That's just lame.  A certain 
language from a certain science fiction media franchise comes to mind 
here (hint: shotgun phonology).

> VI. Really great conlangers
> Okay, this is totally subjective, but just as there are great artists among
> composers and novelists, there are great artists among conlangers. Through
> a combination of innate talent, linguistic-aesthetic sensitivity, and a
> perseverance of effort, they have developed beautiful conlangs whose
> phonology and grammar are both admirably naturalistic. I can think of only
> two people in this category. One is Tolkien; the other is living and has
> gone pro. But even if you’re not Mozart it doesn’t mean you can’t enjoy or
> create music.

I think I can guess whom you mean by "the other is living and has gone 
pro".  But I think there are more than just those two in this category!  
I have seen many *very good* and beautiful naturalistic conlangs from a 
number of authors.  I can't think of any who clearly surpasses Tolkien; 
but there are a few who are at least on a par with him.

> VII. How to create a naturalistic grammar for your conlang
> Let’s suppose you want to have a naturalistic grammar for your new conlang.
> How could you go about it? Here’s a recipe that I’m pretty sure will work
> (though there’s surely other ways to bake this cake):
> 1. Study the grammars of natural languages and choose grammatical features
> that you especially like. Make a list.

I can confirm that.  Designing a naturalistic language requires a good 
idea of what goes in natural languages and what doesn't.  You should 
study grammars of at least a dozen natural languages from all parts of 
the world.

> 2. Eliminate inconsistencies. If you like the head-final and SOV of
> Japanese, but also the case affixes and free-form word order of Latin,
> you’ve got to choose one.

There is some truth in that, but it must be qualified.  Sure, not all 
combinations of typological traits are equally likely.  If in doubt, ask 
WALS (  For instance, there don't seem to be many 
languages that are neither accusative nor SOV.  But "rare" is not 
"impossible", and there are so many perfectly workable feature 
combinations that they cannot all be manifested in the few thousand 
languages that are spoken on Earth.  So why not implement a combination 
of features that is rare?  It may result in something unique!  (My main 
conlang Old Albic is indeed neither accusative nor SOV, by the way.)

> 3. Cross off 40% of the remaining features. Almost for sure, you have too
> many features on your list. We wouldn’t be conlangers if we didn’t like
> interesting language features. Take out 2/5 of your features (or at least
> 1/3). You can incorporate them in another conlang.

Sure - you need to choose a few of the features you have seen, otherwise 
you get a kitchen sink conlang.  Keep the features you can't use in this 
conlang for the next projects.

> 4. Now, build and describe your core grammar.
> 5. Sprinkle with a few irregularities for that diachronic zest. Everyone
> savors a few suppletions and portmanteaus. Don’t over-spice.

The best way of coming up with naturalistic irregularities is the 
diachronic method - create a language and age it by applying sound 
changes and grammatical and semantic changes of the kind that is found 
in natlangs.  Of course, you need a point to start, so if you don't want 
to add another branch to an existing family, you have to sew your 
ancestor language from whole cloth.

> 6. Use your conlang. Speak in it. Write in it. If it doesn’t feel right,
> then reflect on what went wrong and try again.

Another good recommendation.  Use the language, add what is missing (you 
will *always* find things that your grammar doesn't cover adequately) 
and change what doesn't feel right.

> And so, what I am saying here is that if you incorporate a grammatical
> feature that isn’t present in any natlang, then you probably have
> destroyed, or compromised, the naturalism of the grammar of your conlang.

There are two problems with this.  First, you can never know that the 
feature isn't present in any natlang.  Of the 6000+ languages spoken on 
this planet, most aren't adequately documented yet.  There may be a 
language in Brazil or New Guinea or wherever that has the feature you 
are thinking about.  Second, you can creatively extrapolate from things 
that are present in natlangs.  I haven't seen a natlang yet with a 
degree-of-volition marking system of the kind I have in Old Albic; I 
cannot think of any where vowels are autosegmental.  Yet, I think that 
both features of Old Albic are naturalistic, because they are just 
extrapolations from what I have found in natlangs.  The degrees of 
volition came to my mind when I was exploring morphosyntactic alignments 
and hit upon fluid-S languages; the autosegmental behaviour of vowels is 
modelled on the autosegmental behaviour of tones in some languages of 

> If your objective is to engineer a really unusual grammar with heretofore
> unseen features, that’s fine, but if you wish to produce a naturalistic
> grammar, then it’s more prudent to stick with the tried-and-true features
> of natlang grammars. This belief is based on my intuition (admittedly
> unfounded) that the corpus of the world’s natural languages contain among
> themselves virtually every possible natural grammatical feature of human
> language.

I think the latter sentence is not true, at least not literally true.  
WALS lists more than 100 features, each with at least two values.  That 
are more than 2^100 logically possible feature combinations.  Even if 
one removes those that are inconsistent or at odds with well-established 
typological rules, one is left with many millions of possible feature 
combinations - and there are only about 6,000 languages.  There must be 
many perfectly plausible feature combinations which are not realized in 
any natural language.

... brought to you by the Weeping Elf
"Bêsel asa Éam, a Éam atha cvanthal a cvanth atha Éamal." - SiM 1:1