New CBB

Did you get some interesting linguistics thought experiments ideas last time?

Let's get some bunch of newborn childen and move it to an izolated space, probably most suitable for the experiment would be an uninhabitated Earth-like planet. Let's make a conlang with very unnaturalistic or rare features, like trial number or only voiced consonants or three degrees of aspiration or anything else. Then we will learn them only this conlang and when they'll grow up, we'll leave them and let them breed. Do you think unnaturalistic features will disappear? If yes, how long will it last?

_________________
Native
Mutually intelligibile with native language
Almost fluent
Little
Little more
Everybody can speak it!
Speedlang

Any unlearnable feature will disappear in a single generation, or at most in a single lifetime.
As for other features, the less learnable it is, the more vulnerable it will be to rapid change, and the more rapid will be the change to which it is vulnerable.

_________________
I am not responsible for the accuracy of my sources; they're responsible for their own mistakes, if any, and also responsible for defending their own statements if you disagree with them.

To do an extremely agglutinative language with 1-per-1 syllable/morpheme ratio.

I thought about an inclusive but not too sensitive phonology and ended with something like nucleus rules.
It was:
{/i y ɯ u/ or /0 i̯ u̯/ + /e ø ɤ o a ɒ/ + /0 i̯ u̯/} + {long or short}
that makes 116 possible nuclei.
I also started thinking on consonants, but then went to bed.

_________________
(native)
(learning)

--
Języki sztuczne i lingwistyka, po polsku.

What is naturalistic? I suppose most linguistic universals are universal cos there happens to be no counterexamples, not because they were impossible. (It's a philosophic question if you can say that something is (im)possible if it does't exist, though.)

If there are easier ways to express the same thing, yes.
But, if we have made a very good conlang and its odd features are plausible in its context, they bear enough semantics, I think they won't dissappear.

That way of posing the question implies that there are only ways to explain linguistic universals; (1) that a certain feature is logically necessary, or (2) that it just happens to be the case that all languages has that feature.

Explanation (1) fails - at least for many universals - because those universals are statistical rather than truly universal. Explanation (2) also fails, because it would be highly improbable that, for example, 98% of all languages just happened to share a feature.

Thus other kinds of explanations are needed to account for the fact that many linguistic features, though not truly universal, occur among the world's languages with a higher-than-chance frequency.

Suppose languages could have either feature A or feature B, but that feature A is vastly more common.

This could be explained in the following ways:

(3) While both A and B are logically possible, A may be easier to process or something.
(4) While both A and B are logically possible, and also equally easy to process, B only arises given certain very specific conditions. A would be some kind of default, which all languages have on their deepest level of grammar. B may occur as a surface realisation due to certain transformational rules, that work only under certain specific conditions.

_________________
constructedlanguages.net

What you said, (at least mostly), but, additionally:
One constraint that applies to all human natlangs is that they must be learnable within two years by someone who has no language in common with the speech-community whose language s/he is trying to learn. They must be learnable by example and imitation and observation.

---

"Implicational universals" (such as "languages with feature A always or never or almost always or almost never have feature B") are much likelier IMO and IME to be based on logic. If they are statistical rather than absolute, they may be based on practicality.

For instance suppose you want a VSO language that's topic-prominent. Try to design one yourself; you'll see that it's a bit forced. Now try to design a believable diachronic history for it; you'll see that it's very hard to make it believable. Once you've done that it won't surprise you to learn that awfully few natlangs are both VSO and topic-prominent.

"Implicational universals" where the protasis is about more than one feature (such as "Languages that have feature A and also have feature B, always (or nearly always) also have feature C" where A and B and C are pairwise logically (or even statistically) independent), are even more likely to be based on logic and/or practicality. Sometimes the logic isn't obvious at first glance. Conlangers, especially those who construct diachronic language-histories for their conlangs, have a great tool available for finding out what this "hidden logics" are.

IMO, you should not only test by constructing a diachronics, but also by constructing a learnability-path.

_________________
I am not responsible for the accuracy of my sources; they're responsible for their own mistakes, if any, and also responsible for defending their own statements if you disagree with them.