Random ideas: Morphosyntax

[Creyeditor]

I guess German is not like that

---

Reading about a conlang on Reddit, I was just thinking about a language with obligatory quantifier float. Quantifiers would agree in gender, number, person with their head noun and obligatorily occur in initial or post-verbal position, the language being SVO.

Let's start with a sinple SVO sentence.

Kuc gig nak.
cat bite child
'The cat bit the child.'

Now if there is a plural quantifier in the initial position and a plural object, they belong together

Sem-be kuc gig nak-be.
all-PL cat bite child-PL
'The cat bit all the children.'

If there is a post-verbal quantifier in the plural and the only plural argument is the subject, the quantifier is linked to the subject.

Kuc-be gig sem-be nak.
cat-PL bite all-PL child
All the cats bit the child.

Quantifiers would be banned from occuring next to their head noun.

[Salmoneus]

One of my old conlangs had a weird system where the word order was SVOP, with the P being one of a large array of particles marking various transitive sentences of an intransitive verb...it might work something like sa nōki "I speak" > sa nōki la su "I insult you," sa nōki la īni "I tell you," etc. I don't remember much. I recorded that language on Youtube (youtube.com/@Worldquill, iirc), but it was very bad. Very, very bad. I might try the idea again sometime, though.

That sounds a lot like English! Modern English of course uses this less productively than older forms of the language - we mostly use the particles/prepositions just to increase valency, with relatively few 'minimal pairs' where meaning is significantly changed; but in Proto-Germanic, Old English and to some extent AIUI Middle English these pairs were much more common. As Creyeditor points out, it's still a big thing in German. And such minimal pairs do exist in English too - you put the clothes on, but you put the task off. She looked you up, he looked you over. And although we usually use all three of our systems of this interchangeably - verb-object-particle, verb-preposition-indirect-object, and the surviving cases of particle_verb-object - we can even use them contrastively. So looking you over isn't the same as overlooking you. And saying you got off the horse is VERY different from saying you got the horse off...

More broadly, this is an ancestral feature of PIE in general, which manifests differently in different languages. A huge amount of English vocabulary, for instance, is drawn from Latin verbs with these particles attached. [propose, impose, depose, repose, compose, oppose, superpose, etc]

[verb movement is a big part of the differences. From an original SOPV order, when the verbs moved forward in Latin they tended to take the particle with them, and these particles, stranded from their objects but adjacent to the verb, became fused to the verb. In Germanic, sometimes they took the particle with them and sometimes they didn't. In German, they mostly didn't, hence separable verbs (where the particle only fuses when the verb has not been fronted). In English, the particle tended to come with the verb when it was one of a group of one-syllable particles that fossilised (modern a-, be-, for-, maybe a couple of others), or with certain verbs that were frequently paired with particles (eg understand, withstand), and for a while continued to do so productively with 'over' and 'under', but all the other particles tended to be stranded when the verb moved forward - in transitive verbs this lead to SVOP order, while intransitives in SVP turned into valency-increasing SVPO verbs.]

This doesn't mean your idea is bad, of course - English and German and so on work quite well as languages, so there's no reason not to copy them!

[obviously most (all?) of these particles in PIE don't come from serial verbs or prepositional verbs, but rather from rather nebulous 'particles' with a general postpositional/adverbial use. But you're right that verbs are an equally if not more viable source]

[Arayaz]

This doesn't mean your idea is bad, of course - English and German and so on work quite well as languages, so there's no reason not to copy them!

I was aware that some verbs work like this in English, but the idea here is that all verbs would work this way, and the final particles would be particles only, not adpositions ─ and this would be the only way to mark a transitive verb.

[Salmoneus]

I guess German is not like that

---

Reading about a conlang on Reddit, I was just thinking about a language with obligatory quantifier float. Quantifiers would agree in gender, number, person with their head noun and obligatorily occur in initial or post-verbal position, the language being SVO.

Let's start with a sinple SVO sentence.

Kuc gig nak.
cat bite child
'The cat bit the child.'

Now if there is a plural quantifier in the initial position and a plural object, they belong together

Sem-be kuc gig nak-be.
all-PL cat bite child-PL
'The cat bit all the children.'

If there is a post-verbal quantifier in the plural and the only plural argument is the subject, the quantifier is linked to the subject.

Kuc-be gig sem-be nak.
cat-PL bite all-PL child
All the cats bit the child.

Quantifiers would be banned from occuring next to their head noun.

This mostly seems neat and realistic - it reminds me of the gravitation of particles to second position in IE languages, although of course in some ways it's the opposite, as Wackernagelian processes effect unmarked particles, rather than those with actual agreement on them. But nonetheless it seems believable - although I'd expect the agreement to be worn away over time as it get reinterpreted as an adverb.

The "banned from occuring next to their head noun" bit sounds unrealistic to me, though. What would the motivation be? You'd need to have one rule to move the qualifier to the right position - which I think would be fine - but then another rule that a noun couldn't be moved to a certain location if its qualifier was there, or that the qualifier would be chased away if its noun came near it, and... I don't understand the motivation for that. It would feel like the speakers being perverse just for the sake of it...
[though feel free to show me a good reason I haven't thought of yet!]

[Salmoneus]

This doesn't mean your idea is bad, of course - English and German and so on work quite well as languages, so there's no reason not to copy them!

I was aware that some verbs work like this in English, but the idea here is that all verbs would work this way

This less realistic to me, and also less interesting - 'all' and 'always' tend not to be common words in language descriptions! In particular, the idea of a parent language in which there are no transitive verbs feels more like an auxlang idea than a real language idea. I'm not sure I buy the idea of people who, when punched, can only say "that man made a punching motion! And I was nearby and got hurt!", instead of "the fucker punched me!". The transitivity/intransitivity distinction plays a pretty fundamental role in both syntax and pragmatics in most if not all languages. And of course why go to the trouble of justifying an entirely intransitive language, only to immediately reintroduce transitivity through another method? What would happen much more often is that one transitivity marker would become less used while another was becoming more used, leaving a confusing quilt of transitivity methods varying lexically and/or semantically and/or pragmatically.

[of course, there's nothing unrealistic in a language in which a larger proportion of verbs are inherently intransitive; I'm just skeptical of 100% of verbs being inherently intransitive]

, and the final particles would be particles only, not adpositions

This seems fair enough, certainly. It's probably true in English, for a few particles. "Out", for instance. There are a few questionable cases where "out" is prepositional*, but these are rare and can probably be regarded as stylistic elisions of "out of" in modern English.

─ and this would be the only way to mark a transitive verb.

Is it the only way to actually mark a transitive in English? I think it may be. But I assume you mean just the only way to make a verb transitive, whereas in English you can do that just through zero-derivation.



*The world had thrust us from its heart
And God from out his care
And the iron gin that waits for Sin
Had caught us in its snare.
...but poetry isn't the best guide to standard syntax. Eg the previous line is "two outcast men we were", which is so unusual that half the versions on the internet give "two outcast men were we" (which itself isn't super-quotidian!), even though "were" clearly has to be final to 'rhyme' with "care" and "snare".

[Creyeditor]

I guess German is not like that

---

Reading about a conlang on Reddit, I was just thinking about a language with obligatory quantifier float. Quantifiers would agree in gender, number, person with their head noun and obligatorily occur in initial or post-verbal position, the language being SVO.

Let's start with a sinple SVO sentence.

Kuc gig nak.
cat bite child
'The cat bit the child.'

Now if there is a plural quantifier in the initial position and a plural object, they belong together

Sem-be kuc gig nak-be.
all-PL cat bite child-PL
'The cat bit all the children.'

If there is a post-verbal quantifier in the plural and the only plural argument is the subject, the quantifier is linked to the subject.

Kuc-be gig sem-be nak.
cat-PL bite all-PL child
All the cats bit the child.

Quantifiers would be banned from occuring next to their head noun.

This mostly seems neat and realistic - it reminds me of the gravitation of particles to second position in IE languages, although of course in some ways it's the opposite, as Wackernagelian processes effect unmarked particles, rather than those with actual agreement on them. But nonetheless it seems believable - although I'd expect the agreement to be worn away over time as it get reinterpreted as an adverb.

The "banned from occuring next to their head noun" bit sounds unrealistic to me, though. What would the motivation be? You'd need to have one rule to move the qualifier to the right position - which I think would be fine - but then another rule that a noun couldn't be moved to a certain location if its qualifier was there, or that the qualifier would be chased away if its noun came near it, and... I don't understand the motivation for that. It would feel like the speakers being perverse just for the sake of it...
[though feel free to show me a good reason I haven't thought of yet!]

I think both points are valid and maybe show my engelanging side coming through. And both are related. First, the idea of agreement was that you could correctly identify the head noun. Of course, real languages use context cues and strictly speaking, do not need agreement morphology to link quantifier and head.
Second, the non-adjacency requirement. I was thinking of this kind of in terms of stuff like wh-movement, but then I thought I want a fixed word order and consistent positions for subject and object quantifiers. So, initial position for object and post-verbal for subject is kind of perverse, I have to agree. It also makes the agreement uneccesary. I don't think I can really justify that. On the other hand, quantifier float is one of the phenomena that has not been really looked at in detail in many non-SAE languages, so maybe time will tell. In Balinese, quantifiers can float and in some contexts obligatorily float, e.g. in non-pivots, under the assumption that their default position is prenominal. Of course, they can in principle occur directly postnominal in these contexts, so it is not an example of a non-adjacency requirement at all.

[Salmoneus]

Perhaps you could find some other reasoning that would coincidentally arrive at the same point?

For instance, in this simple two-argument-one-qualifier context, you could produce a non-adjacency rule from a V2 rule if you used a different word order:

1. default S precedes O
2. V is always 2
3. Qualifiers float as far forward as they can
4. S is default topic
5. O can be topicalised by fronting
6. Qualifiers apply to the topic

This gives:
cat bite child
all* bite cat* child [Q takes 1, V still in 2 so S drops to 3, but still precedes O so all is well]
child* bite all* cat [O takes 1 (so that it is topic and can be qualified), V still in 2, so Q has to drop to 3, leaving S in 4]

----------

Another way to do it with Somali-style obligatory invisible clefting!

1. all main clauses are VSO
2. Q precedes head
3. all sentences require clefting of focus with zero copula
4. Q is not extracted with head
5. only focus can be qualified

bite cats child > cat bite child <<< [cats [it is that] bite the child]
bite all* cats* child > cat* bite all* child <<< [cats [it is that] all [of which] bite the child]
bit cats* all* children > children* bite cats all* <<< [children [it is that] cats bite all [of which]]

This would even perhaps allow the qualifier itself to be extracted, provided that pivot rules assume that subject co-refer:

bit all* cats* child > all* bite cats child <<< [all [it is, among] cats [that] bite the child]

-------------

oooh, how about this one?

1. VSO
2. possessors precede nouns
3. valency reduction is possible via nominalisation with existential copula ["the dog is eaten" = is dog's eating] (dummy subjects not required)
4. Q must modify S
5. Q always floats to beginning of clause
6. agents of passives expressed through prepositional phrase after verb
7. identificatory copulas exist: "is Bob" (outside of cases of pointing at people) means "it is Bob [that this is true of]" or "it is Bob [that I mean]"
8. optional topicalisation by extraction and fronting
9. objects cannot be extracted
10. clauses can and should elide subjects when the subject is the topic; this can form chains of clauses with shared subjects
11. possessors can also be elided when the possessor is the topic
11. no elision of arguments or of possessors is allowed other than be topicalisation/extraction or by co-reference with topic
12. clauses cannot be joined except by co-reference and elision
12. the copula is zero
13. the preposition governing agents of passives is zero
14. derivation of verbal noun is zero-marked
15. no overt intonation required for well-formed short clause chains



Thus:

bite all* cats* children
> all* bite cats* children [simply by 5]

bite cats all* children*
> Q must modify subject
> new clause required with 'children' as subject: bite cats children, is all* children*
> these clauses cannot be joined: co-reference required
> co-reference impossible, no explicit topic
> topic required: "children: bite cats children, is all children"
> constitutes illegal extraction of object (alternatively: would require illegal elision of object)
> detransitivisation required to make object accessible: "children: is children's biting by cats, is all children"
> elision of co-referential possessor and subject: "children: is biting by cats, is all"
> Q floats to beginning of clause: "children: is biting by cats, all is"
> zero preposition: "children: is biting cats, all is"
> zero-marked verbal noun: "children: is bite cats, all is"
> zero copula: "children: bite cats, all"
> no overt intonation marking: "children* bite cats all*"

OK, so that's a little convoluted, but I think it kind of makes sense?

-------

Anyway, just a few of the more obvious suggestions!

[Creyeditor]

These look like good options. I like all three of them. I think I am emotionally most attached to the second option because it kind of expands on the idea that syntactic alternations that are optional in some natlang become obligatory restrictions in this conlang.

[Acipencer]

(Likely ANADEW)

I was thinking about case marking and I had an idea. Picture a language that marks case with clitics and commonly omits arguments that are obvious through context. The idea is, what if when you omit a noun, you are required to keep its case marking even though it is omitted. As an example, lets say azda means 'see', kozi means 'sheep', and kə/k is the accusative clitic. With this, lets make a basic statement.

kozikə avda
kozi=kə avda
sheep=ACC see
'(I) see a sheep'

Okay, now instead imagine someone just asked you if you can see any sheep right now. The 'sheep' is obvious from context and will likely be dropped. the Clitic, however, stays and attaches to the nearest (and only) other word in the sentence. Now we have:

kavda
k=avda
ACC=see
'(I) see (a sheep)'

[Knox Adjacent]

I guess it makes sense if it originates from a case-marked pronoun in apposition to a noun.

[Omzinesý]

(Likely ANADEW)

I was thinking about case marking and I had an idea. Picture a language that marks case with clitics and commonly omits arguments that are obvious through context. The idea is, what if when you omit a noun, you are required to keep its case marking even though it is omitted. As an example, lets say azda means 'see', kozi means 'sheep', and kə/k is the accusative clitic. With this, lets make a basic statement.

kozikə avda
kozi=kə avda
sheep=ACC see
'(I) see a sheep'

Okay, now instead imagine someone just asked you if you can see any sheep right now. The 'sheep' is obvious from context and will likely be dropped. the Clitic, however, stays and attaches to the nearest (and only) other word in the sentence. Now we have:

kavda
k=avda
ACC=see
'(I) see (a sheep)'

I like that idea! The same morpheme boxes can be used in different buildings.
That's not extremely far from preposition stranding.

[chris_notts]

I've been working or reworking a phonology with apocope and syncope, and I was considering how it might be interesting to add ditropic case clitics into it. A ditropic clitic is a clitic that attach one way phonologically but another way syntactically. For example, Belep has case clitics which attach to the word before the noun phrase whose case is being marked:

ja=me pan=e teâmaa
1PL.INCL.SUBJ=IRR go.TV=DAT high.chief
"We will go to the chieftain"

Note that dative =e is attached to the preceding verb despite the fact it governs the noun teâmaa 'high chief' that follows it. These are not applicatives because the enclitic can be hosted by almost any preceding word, including other arguments (NPs).

Anyway, some languages with apocope and syncope seem to have enclitics which shield or block apocope and otherwise behave as part of the word for vowel deletion processes. Aguaruna is such a language, at least according to the grammar of Simon Overall. It has case enclitics which occur at the end of NPs, and which block apocope of the host's final vowel (since the final vowel isn't final anymore after attachment of the enclitic). For example, namakan = namaka-na fish-ACC, with the final vowel of namaka preserved in the accusative.

So if you combine regular vowel deletion processes with ditropic clitics, which are both independently attested in natlangs, you could get some interesting alternations in the preceding word depending on the case marking of the following NP. For example, if we assume apocope followed by a productive syncope rule that deletes the vowel from a second light syllable (CVCV), then you can get alternations like:

kisin <-- form after apocope
kisina
see
"he/she sees"

kisnal tasaq <-- form after apocope and syncope
kisina=li tasaq
see-ERG man
"the man sees him/her/it"

If the mutating word is the verb and you're just comparing VO and VS, it's almost like an inverse or Phillipine voice system, but of course it stops looking like that if you start inserting other elements. E.g. in a VOS clause, the ergative case marker would attach to the O, not to the initial verb.

I don't know how feasible this is. I don't know of a language where it happens, but ditropic clitics are not that common anyway so it could just be a small sample issue.

The other interesting but maybe unattested thing you could do is something like the contiguity marking of Macro-Je and other Amazonian languages via a phrase-final apocope rule. Macro-Je languages often show alternations in prefixes depending on whether a verb, preposition etc. is directly preceded by its argument or not (thus "contiguity marking"), but in the Macro-Je case the alternation is morphological. If you had a rule of apocope that only applied to the phonological phrase final word, for some definition of the phonological phrase, then you could achieve something like case marking by presence or absence of final vowels. Compare:

kisin <-- form after apocope
kisina
see

kisina tasaq <-- no apocope because VO forms a phrase and the V is not final
kisina tasaq
see man.OBJECT
(he/she/it) sees the man

kisin tasaq <-- apocope because VS does not form a phrase for the purpose of the rule
kisina tasaq
see man.SUBJECT
the man sees (him/her/it)

This is basically a linker like the ubiquitous Tagalog ng, only it's just an underlying lexical specified vowel that gets deleted otherwise. I'm less convinced by the second idea than the first, which I am seriously considering for a conlang.

[Omzinesý]

I've been working or reworking a phonology with apocope and syncope, and I was considering how it might be interesting to add ditropic case clitics into it. A ditropic clitic is a clitic that attach one way phonologically but another way syntactically. For example, Belep has case clitics which attach to the word before the noun phrase whose case is being marked:

ja=me pan=e teâmaa
1PL.INCL.SUBJ=IRR go.TV=DAT high.chief
"We will go to the chieftain"

Note that dative =e is attached to the preceding verb despite the fact it governs the noun teâmaa 'high chief' that follows it. These are not applicatives because the enclitic can be hosted by almost any preceding word, including other arguments (NPs).

Anyway, some languages with apocope and syncope seem to have enclitics which shield or block apocope and otherwise behave as part of the word for vowel deletion processes. Aguaruna is such a language, at least according to the grammar of Simon Overall. It has case enclitics which occur at the end of NPs, and which block apocope of the host's final vowel (since the final vowel isn't final anymore after attachment of the enclitic). For example, namakan = namaka-na fish-ACC, with the final vowel of namaka preserved in the accusative.

So if you combine regular vowel deletion processes with ditropic clitics, which are both independently attested in natlangs, you could get some interesting alternations in the preceding word depending on the case marking of the following NP. For example, if we assume apocope followed by a productive syncope rule that deletes the vowel from a second light syllable (CVCV), then you can get alternations like:

kisin <-- form after apocope
kisina
see
"he/she sees"

kisnal tasaq <-- form after apocope and syncope
kisina=li tasaq
see-ERG man
"the man sees him/her/it"

If the mutating word is the verb and you're just comparing VO and VS, it's almost like an inverse or Phillipine voice system, but of course it stops looking like that if you start inserting other elements. E.g. in a VOS clause, the ergative case marker would attach to the O, not to the initial verb.

I don't know how feasible this is. I don't know of a language where it happens, but ditropic clitics are not that common anyway so it could just be a small sample issue.

The other interesting but maybe unattested thing you could do is something like the contiguity marking of Macro-Je and other Amazonian languages via a phrase-final apocope rule. Macro-Je languages often show alternations in prefixes depending on whether a verb, preposition etc. is directly preceded by its argument or not (thus "contiguity marking"), but in the Macro-Je case the alternation is morphological. If you had a rule of apocope that only applied to the phonological phrase final word, for some definition of the phonological phrase, then you could achieve something like case marking by presence or absence of final vowels. Compare:

kisin <-- form after apocope
kisina
see

kisina tasaq <-- no apocope because VO forms a phrase and the V is not final
kisina tasaq
see man.OBJECT
(he/she/it) sees the man

kisin tasaq <-- apocope because VS does not form a phrase for the purpose of the rule
kisina tasaq
see man.SUBJECT
the man sees (him/her/it)

This is basically a linker like the ubiquitous Tagalog ng, only it's just an underlying lexical specified vowel that gets deleted otherwise. I'm less convinced by the second idea than the first, which I am seriously considering for a conlang.

Interesting!
It's often a problem that, if you don't wanna have cases, nouns have no inflection. Having complex alternations in words that host other word's inflectional marking is a solution.
I have seen ditropic clitics in examples but never payed much attention to them. The term is new to me.

[Creyeditor]

I've been working or reworking a phonology with apocope and syncope

This sounds like a really cool morphophonology idea. I have been thinking about radical syncope languages for some time (or at some point in the past) and I like how you get radically different allomorphs if you combine syncope in even (or odd)-numbered syllables with other phonological processes. I never though about the combination with ditropic clitics before though. I think both your options seem naturalistic. The first option is transparently composed of stuff that independently occurs and I think the second option is, too. Phrase-final vowel deletion is a thing, I think. And phrasing subjects and objects differently is definitely also a thing. Maybe this could be related to Bantu conjoint/disjoint forms.

[Salmoneus]

I've been working or reworking a phonology with apocope and syncope, and I was considering how it might be interesting to add ditropic case clitics into it. A ditropic clitic is a clitic that attach one way phonologically but another way syntactically. For example, Belep has case clitics which attach to the word before the noun phrase whose case is being marked:

ja=me pan=e teâmaa
1PL.INCL.SUBJ=IRR go.TV=DAT high.chief
"We will go to the chieftain"

Note that dative =e is attached to the preceding verb despite the fact it governs the noun teâmaa 'high chief' that follows it. These are not applicatives because the enclitic can be hosted by almost any preceding word, including other arguments (NPs).

Anyway, some languages with apocope and syncope seem to have enclitics which shield or block apocope and otherwise behave as part of the word for vowel deletion processes. Aguaruna is such a language, at least according to the grammar of Simon Overall. It has case enclitics which occur at the end of NPs, and which block apocope of the host's final vowel (since the final vowel isn't final anymore after attachment of the enclitic). For example, namakan = namaka-na fish-ACC, with the final vowel of namaka preserved in the accusative.

So if you combine regular vowel deletion processes with ditropic clitics, which are both independently attested in natlangs, you could get some interesting alternations in the preceding word depending on the case marking of the following NP. For example, if we assume apocope followed by a productive syncope rule that deletes the vowel from a second light syllable (CVCV), then you can get alternations like:

kisin <-- form after apocope
kisina
see
"he/she sees"

kisnal tasaq <-- form after apocope and syncope
kisina=li tasaq
see-ERG man
"the man sees him/her/it"

If the mutating word is the verb and you're just comparing VO and VS, it's almost like an inverse or Phillipine voice system, but of course it stops looking like that if you start inserting other elements. E.g. in a VOS clause, the ergative case marker would attach to the O, not to the initial verb.

I don't know how feasible this is. I don't know of a language where it happens, but ditropic clitics are not that common anyway so it could just be a small sample issue.

The other interesting but maybe unattested thing you could do is something like the contiguity marking of Macro-Je and other Amazonian languages via a phrase-final apocope rule. Macro-Je languages often show alternations in prefixes depending on whether a verb, preposition etc. is directly preceded by its argument or not (thus "contiguity marking"), but in the Macro-Je case the alternation is morphological. If you had a rule of apocope that only applied to the phonological phrase final word, for some definition of the phonological phrase, then you could achieve something like case marking by presence or absence of final vowels. Compare:

kisin <-- form after apocope
kisina
see

kisina tasaq <-- no apocope because VO forms a phrase and the V is not final
kisina tasaq
see man.OBJECT
(he/she/it) sees the man

kisin tasaq <-- apocope because VS does not form a phrase for the purpose of the rule
kisina tasaq
see man.SUBJECT
the man sees (him/her/it)

This is basically a linker like the ubiquitous Tagalog ng, only it's just an underlying lexical specified vowel that gets deleted otherwise. I'm less convinced by the second idea than the first, which I am seriously considering for a conlang.

Both these ideas seem unusual but very plausible to me.

I'm reminded of Proto-Indo-European and early IE languages, where it kind of seems as though word boundaries were relatively weak. Some IE things that spring to mind in this regard:
- the tendency to have chains of clitics in second position, attaching to whatever happens to be in first position
- the tendency for adverbs/adpositions to be welded onto nearby verbs. In Germanic, this welding is still sometimes dependent upon position in the clause
- some analyses of stress patterns in early Latin suggests that stress was phrasal, with preposed enclitics taking away the stress from the main word
- in Irish, clitic pronouns are welded onto adpositions (or vice versa)
- in (Old) Irish, verb complexes are created through the affixation of chains of clitics, and this, through stress shifts to the clitics, as in early Latin, leads to paradigmatic alternations. Hence "berid" (he carries) but "ní-biur" (he does not carry), or "as-berat" (they say) but "ní-epret" (they do not say), or "imm-soi" (he turns around) but "ní-impai" (he does not turn around).
- in Irish, the presence of preposed articles, numerals and pronouns triggers sound changes on the noun, and the noun can trigger sound changes on the following adjective. Specifically, these sound changes are triggered the endings on the preceding words that have already been lost (in a few cases, where the entire preceding word has been lost!)
- similarly in Irish these mutations can be triggered on the second element in a compound, whether or not it's written as one word, and on genitive modifiers. That is, the presence of mutation demonstrates that the two words are bound closely together within a phrase.


These are obviously different from your examples (and mostly concern the left edge of words, not the right), but are close enough that I'd be comfortable with what you're doing!

[Salmoneus]

I've been working or reworking a phonology with apocope and syncope

This sounds like a really cool morphophonology idea. I have been thinking about radical syncope languages for some time (or at some point in the past) and I like how you get radically different allomorphs if you combine syncope in even (or odd)-numbered syllables with other phonological processes.

Yes, I've toyed with this several times, and the real-world (insane) example is Old Irish. But I've always backed away from on it on the grounds of "oh fuck no this is impossible to work with"...

[Creyeditor]

Of course, neither did I succeed in developing this idea any further.

[chris_notts]

I've been working or reworking a phonology with apocope and syncope

This sounds like a really cool morphophonology idea. I have been thinking about radical syncope languages for some time (or at some point in the past) and I like how you get radically different allomorphs if you combine syncope in even (or odd)-numbered syllables with other phonological processes.

Yes, I've toyed with this several times, and the real-world (insane) example is Old Irish. But I've always backed away from on it on the grounds of "oh fuck no this is impossible to work with"...

I've been pondering further and there are some ways to have extensive syncope but also maybe make it manageable, I think. One thing that would help with complex verb morphology is to make the grammatical verbal word be composed of multiple phonological words, which means that there are certain "reset" points which alternate less. Phonological boundaries within the verb word, even in the presence of rigid ordering, aren't uncommon in very synthetic languages, as in Marind, various Athabaskan languages, etc.

[Salmoneus]

And, indeed, in Old Irish, where the verb complex has more and less essentially attached elements, depending on when they became attached, and certain sound changes only effect some and not others.

You can also just make sure that alternations, while dramatic, are predictable and easily explaining.

Or you can do what I'm thinking of doing if I ever do that language up properly - restrict alternations to a small core of irregular commonplace words, and have most of the vocabulary become regular through analogy.

[chris_notts]

And, indeed, in Old Irish, where the verb complex has more and less essentially attached elements, depending on when they became attached, and certain sound changes only effect some and not others.

You can also just make sure that alternations, while dramatic, are predictable and easily explaining.

Or you can do what I'm thinking of doing if I ever do that language up properly - restrict alternations to a small core of irregular commonplace words, and have most of the vocabulary become regular through analogy.

I've just been working through all the forms for the auxiliary/prefix chains given the syncope rules to make sure there isn't too much ambiguity, and there's some fun ones like this:

ē-uqa-jo-ŋen-о̄q > ējuqajoŋenо̄q > ējuqjoŋnо̄q > ējuqqonnо̄q
REFL/MID-lest-VENITIVE-2/3PL-NEG
"Lest they not come and ... themselves"

There's 2 syncopated vowels + cluster assimilation in this case. I've now tuned the forms so the clashes aren't too horrific, but there's still some interesting alternations. Auxiliary forms with many underlying syllables often tend to end up with a geminate or two in them, but I think it's fine since there's plenty of simple high frequency forms too.