Triconsonantal roots and allophony

[jal]

The other day I was wondering to what extent triconsonantal root languages allow allophony to change the root consonants, and whether there are any languages that historical used to be 3cons but due to historic sound changes have obscured this beyond repair. E.g. a first stage could be (abusing Arabian):

aktub (I write, M) > [attup]
kitāb (book) > [tʃitap]
katib (writer, M) > [katʃip]
maktabat (library) > [mattawat]

When it's allophonic rules, speakers would probably still recognize the roots, but a few centuries of further sound changes would freeze them and the system would cease to be productive.


JAL

[KathTheDragon]

As I understand it, the tricon root system exerts massive amounts of analogy on the results of sound changes in order to keep the system intact.

[jal]

As I understand it, the tricon root system exerts massive amounts of analogy on the results of sound changes in order to keep the system intact.

That's what I would expect, and what can be seen in Semitic languages, afaik. Thanks for confirming this. I'm wondering though if there are counter examples, but I wouldn't be surprised if there weren't.


JAL

[vec]

Which may explain why we're able to trace Afro-Asiatic languages back much further than any other family, really.

https://en.wikipedia.org/wiki/Afroasiatic

[Yng]

It is the case with Arabic at least that there are oddly few cases of conditional sound changes, although one of the changes you posit, the affrication of /k/ before front vowels, has happened in some dialects - I don't believe this has resulted in the reanalysis of roots in this case, but rather the modification of one of the consonants where it appears in certain ablauts (so you could have for example čitāb, pl. kutub). Likewise for the velarisation of some /r/, which happens in certain positions in certain ablauts, and Hebrew/Aramaic begadkefat (frication of stops postvocalically). The Semitic system generally seems to lend itself very well to analogising away conditional changes, though (and this is probably helped by reference to the omnipresent Classical Arabic).

That said, I think you're all misunderstanding the nature of triconsonantal languages. You seem to be imagining that roots are a magic, constant, platonic force which are inserted into patterns to produce words, which is the traditional analysis but creates a kind of Semitic exceptionalism. Roots are, to some extent, a thing - but Arabic is perfectly happy to extract new reanalysed roots from words, including borrowings and existing words already in patterns (tamaḥwara < miḥwār < ḥiwār), and words take semantic features from other, related words, not just from the abstract 'meaning of the root' which everyone gets so excited about - maktabah 'library' is surely far more logically a derivation from kitāb than from a weird handwaving of 'place of writing'ˌ for example. This all suggests to me that just like any other language, Arabic's derivational morphology is much more about analogy with the relationship between two existing pieces of vocabulary and applying that relationship, as a process, to another existing piece of vocabulary. Admittedly in Arabic's case this is slightly different from just adding an affix since it involves complicated phonological changes to the input word, including (sometimes) the loss of existing affixes, vowel deletions and transformations and so on - but this doesn't mean it's fundamentally any different. Various different types of input words, including ones with less than three consonants (like dōr), can be modified to fit different ablauts (> dwār, analogous to film > flām).

From another angle, sound changes are often dealt with by analogical jumps in the system, too. Historically glottal stops disappeared, leaving verbs like qaraʔ as qara. Did Arabic speakers get confused and upset and stop using qara? No, they just analogised it to conjugate like existing verbs which looked similar.

Anyway, as far as the original question is concerned - roots don't really exist, so 'roots becoming unrecognisable' is not a problem. The issue would be that relationships between existing words would be obscured (as they are in other languages sometimes), but analogy could deal with this problem conveniently. There's no reason why ablauts couldn't accept processes like consonant mutation/gradation, vowel quality changes, tone shifts and so on - as long as these are generally consistent across the application of a given ablaut.

Anyway TL;DR get rid of the idea of triconsonantal languages in the first place because it's dumb and isn't a good tool for analysis

[Xephyr]

Fwiw, the kind of consonantal assimilation the OP was talking about happens in Biblical Hebrew:

nātan (give.PERF) + tî (1sg) -> nātattî "I gave"

in Syriac:

ezzat "she went", ezzet "I went"; from *ʔZL, cf. ezal "he went"

and in Akkadian:

iqqur "s/he destroyed", taqqur "you(m) destroyed", etc.; from *naqārum "to destroy"

[Salmoneus]

You seem to be imagining that roots are a magic, constant, platonic force which are inserted into patterns to produce words, which is the traditional analysis but creates a kind of Semitic exceptionalism. Roots are, to some extent, a thing... This all suggests to me that just like any other language, Arabic's derivational morphology is much more about analogy with the relationship between two existing pieces of vocabulary and applying that relationship, as a process, to another existing piece of vocabulary. Admittedly in Arabic's case this is slightly different from just adding an affix since it involves complicated phonological changes to the input word, including (sometimes) the loss of existing affixes, vowel deletions and transformations and so on...

Which makes it confusing to say

roots don't really exist

...because you've just described the existence of roots. And you then go on to say

The issue would be that relationships between existing words would be obscured (as they are in other languages sometimes)

which presupposes the idea of roots (i.e. phonetic elements in common systematically between words of related meaning).

Anyway TL;DR get rid of the idea of triconsonantal languages in the first place because it's dumb and isn't a good tool for analysis

That's just ridiculous - a kneejerk response. You've already said yourself that triconsonantal languages are "different", because derivation in them isn't "just adding an affix since it involves complicated phonological changes to the input word, including (sometimes) the loss of existing affixes, vowel deletions and transformations and so on". This is significant: as you say, it makes these languages different. What's different specifically? Well, in these languages the roots generally take the form of three, often-discontinuous, consonants, rather than a single string of consecutive phonemes. Since the distinguishing feature is that the roots are triconsonantal, 'triconsonantal root language' seems a pretty good general label for them (notwithstanding that, of course, not all roots can necessarily be explained in this way in all cases, and also notwithstanding that discontinuous morphological alterations also occur to some extent in some other language).

The only way I can see to justify your view is if you insist on defining roots as some sort of mystical platonic entity. But that makes no sense either: either you believe in mystical platonic entities, in which case you wouldn't object to the classification, or you don't, in which case you wouldn't insist on defining roots that way. It's as though I said "seabirds are the incarnated spirits of the dead" and you angrily retorted "there's no such thing as incarnated spirits of the dead - so stop talking about 'seabirds', there's no such thing!" (except, of course, that nobody in this case has actually claimed that seabirds ARE the incarnated spirits of the dead).

[Khvaragh]

Generally, I think there are different interpretations of what triconsonantal roots are, either as convenient fictions (basically invented by the Arab grammarians/lexicographers as a way to organize dictionaries and, later, analyze derivational morphology - note also that grammatical traditions in Hebrew and Aramaic derive from the Arabic grammatical tradition, which has significant pragmatic repercussions for how Semitic languages are generally understood) - or as a kind of discontinuous bound morpheme. Note also that certain roots have characteristic vowels, which sometimes disambiguate different roots, i.e. ḥalama "he dreamed" vs. ḥaluma "he was mild-tempered," which complicates the picture. This difference tends to be obscured in less conservative Semitic languages like Hebrew and Aramaic.

[احمکي ارش-ھجن]

Generally, I think there are different interpretations of what triconsonantal roots are, either as convenient fictions (basically invented by the Arab grammarians/lexicographers as a way to organize dictionaries and, later, analyze derivational morphology - note also that grammatical traditions in Hebrew and Aramaic derive from the Arabic grammatical tradition, which has significant pragmatic repercussions for how Semitic languages are generally understood) - or as a kind of discontinuous bound morpheme. Note also that certain roots have characteristic vowels, which sometimes disambiguate different roots, i.e. ḥalama "he dreamed" vs. ḥaluma "he was mild-tempered," which complicates the picture. This difference tends to be obscured in less conservative Semitic languages like Hebrew and Aramaic.

What more can you tell me about certain roots having characteristic vowels? I have been wondering why when I look at the Hebrew verb conjugation page on Wikipedia all the vowels are different for each root that they don't even match the binyanim templates. It just seems like this would cause an ambiguation for certain patterns. I also want to make my own 3con root language more realistic, so might give me more insight.

[Khvaragh]

Generally, I think there are different interpretations of what triconsonantal roots are, either as convenient fictions (basically invented by the Arab grammarians/lexicographers as a way to organize dictionaries and, later, analyze derivational morphology - note also that grammatical traditions in Hebrew and Aramaic derive from the Arabic grammatical tradition, which has significant pragmatic repercussions for how Semitic languages are generally understood) - or as a kind of discontinuous bound morpheme. Note also that certain roots have characteristic vowels, which sometimes disambiguate different roots, i.e. ḥalama "he dreamed" vs. ḥaluma "he was mild-tempered," which complicates the picture. This difference tends to be obscured in less conservative Semitic languages like Hebrew and Aramaic.

What more can you tell me about certain roots having characteristic vowels? I have been wondering why when I look at the Hebrew verb conjugation page on Wikipedia all the vowels are different for each root that they don't even match the binyanim templates. It just seems like this would cause an ambiguation for certain patterns. I also want to make my own 3con root language more realistic, so might give me more insight.

This is much less obvious in Hebrew and Aramaic because they have undergone much more sweeping vowel changes and mergers because of ancient stress changes, loss of morphological markers, etc. which still exist in (Classical and Modern Standard) Arabic - viz Arabic kataba vs. Syriac ktab. There's also the fact that particularly in Hebrew, the so-called "gutturals" ח ע ה א ר have caused further vowel changes which is why the binyanim paradigms for verbs containing these consonants are "irregular." AFAIK, the characteristic vowels, already rather vestigial in Classical Arabic - there aren't a lot of roots where the characteristic vowel marks a different root - based on some theories, at an earlier stage of the language, they may have had a morphological function - a medial /a/ for transitive, /u/ and /i/ mostly intransitive - seem to be lost or totally non-productive in Hebrew and Aramaic. The only significant characteristic vowel alternation I can think of is Syriac qām (root consonants q-w-m) "he rose" vs. mit "he died," but, at least according to my Syriac teacher Dr. Thackston, mit is the only -i- medial verb in the whole language (though sām "he placed" has an irregular imperfect, nsim, tsim, etc. which may possibly be a remnant.) There seem to be no doublets though, no *māt. There are initial and final characteristic vowels as well, i.e. bnā "he built" vs. ḥdi "he rejoiced." However, these are all weak roots, so it's difficult to say whether these really have anything to do with the ancient characteristic vowels or they're just the result of weak consonant reduction (to long vowels).

[sirdanilot]

Urm, root consonants 'disappear' in Hebrew all the time, except they don't in random places but these are fairly predictable. However, changing of root consonants (other than very obvious things such as begadkefat) is just not a very good idea since the consonants kind of 'make' the words in Semitic, which already makes them much more 'stable' to begin with.

In Arab, it's mostly the non-root consonants that tend to change (such as t -> T in some roots that already have an emphatic consonant).

[jal]

changing of root consonants is just not a very good idea

Language evolution typically doesn't bother with whether it's a good idea or not, they just happen. Speakers cope.


JAL

[sirdanilot]

If it's not a good idea, it's less likely to happen. Of course it often still happens, but FAITHFULLNESS is also a constraint ! (not sure if I have my optimality theory right here)

[jal]

If it's not a good idea, it's less likely to happen.

I smell a bit of circular reasoning here.


JAL

[sirdanilot]

No, because I haven't explained what I mean by 'not a good idea'.

I mean that once consonants within something ike semitic start changing around, at least, root consonants, then the words become kinda unrecognizable, since the root consonants are the main semantic-information holding cues in these langauges, especially in verbs In nouns, this is not as prevalent since though these are of course formed by processes similar to those in verbs they are usually not analyzed that way (people know that kitab is related to kataba because they learned it, they don't actively form words like kitab from kataba every single time). But with a paradigm like kataba yaktubu, if the latter would become something like *yattuv or something then that gets kinda completely different from the root kitab, and with a productive process like that this just doesn't tend to happen.

And because mostly sound changes tend to apply to the entire language rather than one certain word class the root consonants in these languages tend to stay *mostly* stable.

[احمکي ارش-ھجن]

No, because I haven't explained what I mean by 'not a good idea'.

I mean that once consonants within something ike semitic start changing around, at least, root consonants, then the words become kinda unrecognizable, since the root consonants are the main semantic-information holding cues in these langauges, especially in verbs In nouns, this is not as prevalent since though these are of course formed by processes similar to those in verbs they are usually not analyzed that way (people know that kitab is related to kataba because they learned it, they don't actively form words like kitab from kataba every single time). But with a paradigm like kataba yaktubu, if the latter would become something like *yattuv or something then that gets kinda completely different from the root kitab, and with a productive process like that this just doesn't tend to happen.

And because mostly sound changes tend to apply to the entire language rather than one certain word class the root consonants in these languages tend to stay *mostly* stable.

What if all the word containing the roots changed?

If speakers realized yaktub turned into yattuv, maybe they turn kataba into tatava.

[sirdanilot]

Seems very exotic and unlikely to me, but that is mostly supported by the fact that this never happened anywhere as far as I know. I mean, Hebrew pe-nun verbs (root starts with n, and this n tends to be assimilated to another consonant a lot or 'disappear') still start with n and don't magically lose the n in forms where there's no reason to.

[jal]

I mean that once consonants within something like semitic start changing around, at least, root consonants, then the words become kinda unrecognizable, since the root consonants are the main semantic-information holding cues in these langauges, especially in verbs. In nouns, this is not as prevalent since though these are of course formed by processes similar to those in verbs they are usually not analyzed that way

There are many examples of natural languages with totally weird verbal and/or nominal paradigms that where triggered by historical sound changes. The mental connection between "yaKTuB" en "KaTaBa" because of the root consonants may be lost, but "yattub" and "tatava" are just as plausible as, say, "I am" and "I were". Sure, maybe in the long run all these irregularities may be replaced by new paradigms, but the original tricon roots would be gone.

Seems very exotic and unlikely to me, but that is mostly supported by the fact that this never happened anywhere as far as I know.

The fact that it has never happened is indeed a strong indicator, but on the other hand, there's just (afaik) a single language family that has tricon roots, and maybe there's other constraints resisting these type of sound changes.

I mean, Hebrew pe-nun verbs (root starts with n, and this n tends to be assimilated to another consonant a lot or 'disappear') still start with n and don't magically lose the n in forms where there's no reason to.

"Reason" is not really involved in (natural) language change, generally speaking.


JAL

[sirdanilot]

JAL I am just using terms like ' not a good idea ' and ' reason ' to explain the processes here, I am not saying that sound changes or languages are conscious entities or anything. I would hope that you would be able to abstract away from my laymen's terms here.

Well, of course in Arab loanwoards, any form of triconsonantalness is lost. Swahili kitaba -> vitaba (off the top of my head) and nothing like kutub. But that's not so surprising, reanalysis and all. Someone told me they call a roundabout ' kipilefti' with a plural of ' vipilefti', haven't checked if that's true though !

What annoys me a bit on this forum is that people are trying to see ' triconsonantal root languages' as some sort of separate system, and that they are a bit TOO disappointed with ' the fact that semitic is the only group that has this cool feature, boo hoo !!!' . This is not how it works. Triconsonantalness and Semitic are intrinsically tied together. A non-semitic tri consonantal root system SIMPLY DOES NOT EXIST. It has no use to talk of these kind of systems as some sort of separate entity which only happens to be expressed in Semitic. No, it's a very typically Semitic feature, triconsonantalness and semitic are intrinsically tied. If you want to make a conlang with a SIMILAR system then by all means, but it has very little basis in the real world.

[Ryan of Tinellb]

It seems the point is to evolve a language perhaps towards a triconsonantal-type system, rather than just creating one from whole cloth. I'm looking forward to doing that myself on my current project.

[jal]

It seems the point is to evolve a language perhaps towards a triconsonantal-type system, rather than just creating one from whole cloth. I'm looking forward to doing that myself on my current project.

Or, in my case (hence the OP), to evolve it away from it.


JAL

[Ryan of Tinellb]

Oh, now I get it. Cool. I've gotten so used to seeing people wanting to make them, not break them.

[Astraios]

jal, look at Neo-Aramaic. It has such things as postvocalic /b/>/w/, leading to roots like ˁ-b-d (work) becoming ˀ-w-z and giving synchronic "irregularities" like so:

gēwiz (*k-ˁāβið) 'he does' ~ ˀuzle (*ˁβīð-lēh) 'he did it'
gyātu (*k-yāθiβ) 'he sits' ~ tūle (*yθīβ-lēh) 'he sat'

, and other similar developments leading to things like:

kīˀe (*k-yāðiˁ) 'he knows' ~ zeˀle (*yðīˁ-lēh) 'he knew it'

In spite of this, Neo-Aramaic is still very much triconsonantal (or whatever else Yng wants to call a language whose roots are built around a triconsonantal frame) (though it has reduced the binyanim and replaced the tense system with participials), so it is possible to put up with a suprisingly lot of allophony and irregularity.

It is similar to Hebrew allophony as well:

natati (*natan-tī) 'I gave' ~ ˀeten (*ˀa-ntin) 'I will give'
nafalti (*napal-tī) 'I fell' ~ ˀepol (*ˀa-npul) 'I will fall'
nagati (*nagaˁ-tī) 'I touched' ~ ˀega (*ˀa-ngaˁ) 'I will touch'
baxiti (*bakay-tī) 'I cried' ~ ˀevke (*ˀa-bkay) 'I will cry'

Thanks in part to the conservative orthography, I guess, there isn't much analogising of allophony though:

karati (*qaraˀ-tī) 'I read' ~ ˀekra (*ˀa-qraˀ) 'I will read'
kariti (*qaray-tī) 'I happened' ~ ˀekre (*ˀa-qray) 'I will happen'
kariti (*karay-tī) 'I mined' ~ ˀexre (*ˀa-kray) 'I will mine'

So basically it's still going to take more centuries yet for the triconsonantality to disappear.

AFAIK, the characteristic vowels, already rather vestigial in Classical Arabic - there aren't a lot of roots where the characteristic vowel marks a different root - based on some theories, at an earlier stage of the language, they may have had a morphological function - a medial /a/ for transitive, /u/ and /i/ mostly intransitive - seem to be lost or totally non-productive in Hebrew and Aramaic.

They're not productive, but they're certainly still visible in alterations like:

gadal (*gadala) 'he got bigger' ~ gadol (*gadula) 'he is big' ~ gadel (*gadila) 'he gets bigger'
katan (*qaṭana) 'he is small; he got smaller' ~ katon (*qaṭuna) 'he is unworthy' ~ katen (*qaṭina) 'he gets smaller'

[jal]

jal, look at Neo-Aramaic

Cool, that's basically what I was getting at! Thanks.


JAL