How much can a language change grammatically?

[Isfendil]

Now, I know that this seems like an obvious question, but the specific series of subquestions I ask with it are not obvious- hence why my research for them did not turn anything up.

Some background: The thing is, I am making a Semitic language for a conworld. Before that, I was making an indo European language for this world as well. This world is not ours, and most of its languages are unique but I wanted to both include a way for a language nearly identical to English (hence why another language in the IE family would be present) to appear and also on top of that include my favourite family of western languages in.
Now, the thing is, I don't speak Semitic languages. I can speak 2-3 indo European languages and write four, in three different language subfamilies. I do plan on learning... An unhealthy amount of Semiticlangs in university due to a desire for studying both the Islamic and pre biblical near east, but that's not going to really help me with my questions.

In my IE language, I have invented a lot of new words in addition to the ones I derived from PIE (because the family is original) and the grammar draws inspiration from all of the IE grammars I'm familiar with. I have varieties of conjugations from Romance, have-insertion, some particles, and possssive prefixes from Farsi, the genders have collapsed like in Farsi(at least I think they collapsed in Farsi- or did Anatolian/Iranian never have them?), and SOV (occasional OSV) order, which I just thought sounded cool and didn't take from anywhere in particularly.
The thing is, I also added a passive "tense" and an infinitive/timeless "tense", because it didn't seem out of the question to me, and then I also added the occasional OSV word order for theatrics. These seemed okay to me, as I mentioned, because I felt comfortable enough with this family that I could do so.

So, with all that in mind, just how much can I add/subtract and mutate from my Semitic language before- even though it has all the derived words, and the order- it couldn't be considered Semitic anymore? I know I can't just agglutinate it or remove cases, mess around with roots- but, say I wanted to add a new case? A language that I'm reading about, Akkadian, has no "to be" verb, so they manipulate order and have stem changes to deal with it- but what if I just had a case/conjugation that acted as a "to be-izer" (ex: Zuna Malêk-is, "Zuna is king")? What if I wanted to make a new preposition? Or collapse a pronoun? Or even replace one of the triconsonantal roots/have it deconstruct into to different sub roots?
(As a side digression, where the heck did "šarr" come from!? What about MLK!?)

And phonology? I am pretty sure that it's safe making syllable phonology and sound inventories pretty unique, but are there sounds that a family must have in common? Let me specify that I don't mean the sounds that EVERY language has in common, like nasals and bilabials, but familial ones- I have yet to encounter a Semitic language that doesn't have [x] or [q] in it, for instance.

So, with that in mind, just how much CAN a language change without being disowned by its family, so to speak?

[Creyeditor]

Okay, I think your question is not valid. There is no upper limit to language change. There are, however, two related questions:

• How much can a language change until you cannot recognize its family anymore?
• How fast can a language change?

The first question has a very obvious, yet subjective upper limit. The second question is more difficult.

[Davush]

Maltese has neither /x/ nor /q/. /x/ and /ħ/ merged into /ħ/ and /q/ with /ʔ/. Maltese also has some other quite divergent changes from other varieties of Arabic which may give you some ideas.

Also, modern Arabic dialects do not have case (or very little traces of it), so removing case totally isn't unheard of.

As for the question, I think most language families share a vague set of traits - some of which are usually present in most languages of that family, but internal diversity can still be very high. Naturally, you'd expect there to be a certain amount of cognates in the lexicon (this also depends on loanwords and so on). If your language is a 'direct descendant' of Proto-Semitic, then I'd probably expect it to show less diversity than say, 3-4 'generations' down the line, but even then languages can vary considerably. Think how different the various descendants of PIE are.

[Isfendil]

Also, modern Arabic dialects do not have case (or very little traces of it), so removing case totally isn't unheard of.

What about adding new case, like in the example I mentioned? Grammar changing or expanding? Does that even happen? How can everything be reductive when those ancient Hunter-Gatherer people knew nothing of grammar and were simply ordering the world into patterns?

[GrandPiano]

New cases can definitely arise. As you point out, they wouldn't exist if they couldn't. IIRC, cases generally arise when words merge together. Say you have a postposition "gu" which marks the object of a verb and a noun "emob", so "emob" is used as the subject and "emob gu" is used as the object. Then, over time, people start morphing "gu" into the word before it, so "emob gu" becomes "emoggu". Now "emob" is the nominative case form and "emoggu" is the accusative case form.

[Isfendil]

New cases can definitely arise. As you point out, they wouldn't exist if they couldn't. IIRC, cases generally arise when words merge together. Say you have a postposition "gu" which marks the object of a verb and a noun "emob", so "emob" is used as the subject and "emob gu" is used as the object. Then, over time, people start morphing "gu" into the word before it, so "emob gu" becomes "emoggu". Now "emob" is the nominative case form and "emoggu" is the accusative case form.

And then over generalization finishes the job?

[GrandPiano]

As in, generalization to other nouns?

[Isfendil]

As in, generalization to other nouns?

Yes, exactly, when creating a new case or tense?

[GrandPiano]

Well, in my hypothetical example, the same process would already apply to all other nouns, without the need for any further generalization. Just as "emob gu" would become "emoggu", "fis gu" would become "fizgu", and "ohat gu" would become "ohaggu" (or other things; there are probably a number of possible sound changes that could occur).

I suppose if "gu" attaches to the last word of a noun phrase rather than the head, generalization might change that. For example, say there's an adjective "rina":

emob rina gu > emob rinagu > emoggu rina

This might even be where agreement comes from:

emob rina gu > emob rinagu > emoggu rinagu

[DesEsseintes]

Well, in my hypothetical example, the same process would already apply to all other nouns, without the need for any further generalization. Just as "emob gu" would become "emoggu", "fis gu" would become "fizgu", and "ohat gu" would become "ohaggu" (or other things; there are probably a number of possible sound changes that could occur).

I suppose if "gu" attaches to the last word of a noun phrase rather than the head, generalization might change that. For example, say there's an adjective "rina":

emob rina gu > emob rinagu > emoggu rina

This might even be where agreement comes from:

emob rina gu > emob rinagu > emoggu rinagu

emoggu rinagu sounds so cute!

[Isfendil]

Well, in my hypothetical example, the same process would already apply to all other nouns, without the need for any further generalization. Just as "emob gu" would become "emoggu", "fis gu" would become "fizgu", and "ohat gu" would become "ohaggu" (or other things; there are probably a number of possible sound changes that could occur).

I suppose if "gu" attaches to the last word of a noun phrase rather than the head, generalization might change that. For example, say there's an adjective "rina":

emob rina gu > emob rinagu > emoggu rina

This might even be where agreement comes from:

emob rina gu > emob rinagu > emoggu rinagu

Thank you! This is actually a perfect justification for language production. And here I dreaded the possibility of endless degeneration... I feel a bit more free with the Semitic conlang now, thanks.

emoggu rinagu sounds so cute!

I agree.

[GrandPiano]

I'm pretty sure what I said is correct, anyway. Someone with more experience in diachronics can say whether it is.

[Lao Kou]

emoggu rinagu sounds so cute!

emoggu rinagu - hakuna matata


It's our problem-free philosophy... emoggu rinagu!

[cntrational]

I'm pretty sure what I said is correct, anyway. Someone with more experience in diachronics can say whether it is.

Yeah, it's correct. But to be clear, it usually happens by soundchange, like "all stop clusters merge" causing bg to become gg, then being generalized by analogy.

[Xing]

So, with all that in mind, just how much can I add/subtract and mutate from my Semitic language before- even though it has all the derived words, and the order- it couldn't be considered Semitic anymore? I know I can't just agglutinate it or remove cases, mess around with roots- but, say I wanted to add a new case? A language that I'm reading about, Akkadian, has no "to be" verb, so they manipulate order and have stem changes to deal with it- but what if I just had a case/conjugation that acted as a "to be-izer" (ex: Zuna Malêk-is, "Zuna is king")? What if I wanted to make a new preposition? Or collapse a pronoun? Or even replace one of the triconsonantal roots/have it deconstruct into to different sub roots?
(As a side digression, where the heck did "šarr" come from!? What about MLK!?)

And phonology? I am pretty sure that it's safe making syllable phonology and sound inventories pretty unique, but are there sounds that a family must have in common? Let me specify that I don't mean the sounds that EVERY language has in common, like nasals and bilabials, but familial ones- I have yet to encounter a Semitic language that doesn't have [x] or [q] in it, for instance.

So, with that in mind, just how much CAN a language change without being disowned by its family, so to speak?

The notion of 'being i Semitic language' is, at least in the strict linguistic sense, a genetic one – a language is a Semitic language if it has evolved from a Semitic parent-language through a series of systematic sound changes (accompanied by other changes in grammar, semantics etc.) In that sense, a language cannot really be 'disowned' by its language family. It doesn't have much with surface similarity to do. Given enough time, virtually anything can change into anything in linguistics..., and I don't think any of the changes you have mentioned is obviously 'weird' or implausible. A language would definitely don't stop to be 'Semitic' if, [x] hardened or [q] fronted into [k].

(There are a few complications: certain linguistic features tend to go together, so that if one of them is changed, one might expect the other one to change too. Usually, these are tendencies rather than absolute rules, but if violate very many of these tendencies, you might receive comments about your language being not being 'naturalistic'. For instance, lots of cases tend go together with SOV, ergative alignment tend not to be found in SVO languages, VO languages tend to have lots of agreement, etc.)

One could ask, could Semitic languages (or languages of any other language family), ever diverge to such an extent that it's no longer meaningful to speak of a unified family? No. Since we say that two languages belong to the same family if they have evolved from the same parent-language, this hasn't anything to do with how much the have diverged from the parent-language, or from each other. Now there is a little complication: We should really say, two languages belong to the same family if they have evolved from the same parent-language, as far back as we are able to reconstruct. It's highly possible that many of today's reconstructed proto-languages – PIE, Proto-Uralic, Proto-Afroasiatic (of which Proto-Semitic is a daughter-language) in turn are related is some way, by having evolved from a common parent-language. But we cannot reconstruct such a language (at least not with any degree of certainty; proposed reconstructions are rather speculative). Therefore, we don't say that, for example, English and Arabic are related (though they could really be, if go sufficiently far back in time).

[Curlyjimsam]

In cases of creolisation the grammar of a language can change massively over the course of a few decades. Less extreme but still pretty rapid cases of major language change might occur in contexts where there are very high numbers of second-language speakers, or where a language is pidginised/creolised but undergoes some degree of depidginisation.

You might no longer consider the same language as part of the same family, but the concept of "language family" is basically just a useful descriptor with no clear boundaries in anything in the real world.

[Isfendil]

(There are a few complications: certain linguistic features tend to go together, so that if one of them is changed, one might expect the other one to change too. Usually, these are tendencies rather than absolute rules, but if violate very many of these tendencies, you might receive comments about your language being not being 'naturalistic'. For instance, lots of cases tend go together with SOV, ergative alignment tend not to be found in SVO languages, VO languages tend to have lots of agreement, etc.)

Could you elaborate on this? What changes are actually tied together? Maybe there's a forum post or book...