Random ideas: morphosyntax

[Kuchigakatai]

This is pretty much my favourite thread over on the CBB forum, and I wonder if we could get it going here too. It's just a thread to post random ideas about grammar, whether involving morphology or syntax or both. Here's an example:


I came up with an alignment where
- a single subject argument is typically marked with case #1
- by default, if there's both a subject and object, the subject is marked with case #1, and the object with case #2
- if pragmatic focus falls on the subject, the subject is marked with case #3, and the object with case #1 (!)
- case #2 and case #3 normally do not co-occur

One way I interpret this is that #1 is a "topic" case (TOP), #2 is an accusative (ACC) and #3 is a nominative (NOM).

man-TOP came 'The man came.'
man-TOP knife-ACC grab 'The man grabbed a knife.'
knife-TOP man-NOM grab 'The man grabbed a knife.' (focus on "man": It was the man who grabbed a knife, the knife was grabbed by the man)

And again, "NOM" and "ACC" do not co-occur. When one is used, the other argument is taken up by "TOP" marking. (I suppose I could allow the subject of an intransitive verb to be marked "NOM" if focus falls on the subject noun phrase specifically. man-NOM came.)

I mentioned this idea elsewhere and someone told me Japanese is vaguely like this, as statistically the subject particle =ga and the object particle =o don't co-occur much, and both particles are associated with focus. Which surprised me, but mind you that (in spite of what my username might suggest) I don't know Japanese...

[bradrn]

I could have sworn we had a thread like this a while ago, and it got abandoned? But anyway:

- a single subject argument is typically marked with case #1
- by default, if there's both a subject and object, the subject is marked with case #1, and the object with case #2
- if pragmatic focus falls on the subject, the subject is marked with case #3, and the object with case #1 (!)
- case #2 and case #3 normally do not co-occur

This is just a split-ergative alignment. Cases #2 and #3 are straightforwardly accusative and ergative respectively, with #1 being the unmarked nominative/absolutive. I’m not aware of any language where the alignment split is solely based on focus, but it’s often an important factor, especially in optional-ergative languages: Tibetan is the usual example here. Such systems often (or maybe always?) allow intransitive subjects to take #3 as well, especially when they’re agentive or focussed — the standout case here is probably Ma Manda, which combines topic-prominence with an ergative system which is practically marked-nominative given the behaviour of S (Pennington 2013).

[Kuchigakatai]

I could have sworn we had a thread like this a while ago, and it got abandoned? But anyway:

- a single subject argument is typically marked with case #1
- by default, if there's both a subject and object, the subject is marked with case #1, and the object with case #2
- if pragmatic focus falls on the subject, the subject is marked with case #3, and the object with case #1 (!)
- case #2 and case #3 normally do not co-occur

This is just a split-ergative alignment. Cases #2 and #3 are straightforwardly accusative and ergative respectively, with #1 being the unmarked nominative/absolutive. I’m not aware of any language where the alignment split is solely based on focus, but it’s often an important factor, especially in optional-ergative languages: Tibetan is the usual example here. Such systems often (or maybe always?) allow intransitive subjects to take #3 as well, especially when they’re agentive or focussed — the standout case here is probably Ma Manda, which combines topic-prominence with an ergative system which is practically marked-nominative given the behaviour of S (Pennington 2013).

Hmm...

man-ABS come
man-ABS knife-ACC grab
knife-ABS man-ERG grab

I see what you mean, and I agree it's split-ergative.

The link you gave for Pennington (2013) doesn't work, but this one does, and here's the accompanying PowerPoint presentation. I'll check out this Ma Manda stuff, thanks!

I still find the vague similarity with Japanese interesting though. Maybe I could make the conlang evolve from something Japanese-like to something split-ergative or viceversa.

[bradrn]

The link you gave for Pennington (2013) doesn't work, but this one does

Yep, that’s the one!

and here's the accompanying PowerPoint presentation

…but I had no idea this existed! He also has a pretty good reference grammar of the language.

I still find the vague similarity with Japanese interesting though.

Japanese is basically the nominative–accusative counterpart of Ma Manda, alignment-wise. I believe it has to do with the semantics of topic-marking: nominative and ergative arguments tend to be agentive, and ergative case-marking in particular often gets used for focussed subjects, whereas topics tend to be non-agentive, with topic and focus roles being mutually exclusive.

Maybe I could make the conlang evolve from something Japanese-like to something split-ergative or viceversa.

‘vice versa’ is basically what happened to Ma Manda: it used to be optionally ergative, but the ergative marker got generalised to intransitive subjects, and is now obligatory on some of them, making it marked nominative.

[chris_notts]

Cross-posting something I submitted on CBB about a real feature of some Mayan languages which I thought was cool enough to steal for my latest conlang.

In some Mayan languages, deictic distinctions (this/that, here/there, …) are not marked by individual words or morphemes but are split between a distance neutral deictic trigger word which determines the function (presentative, adverbial, adnominal, …) and a series of clause final particles (not necessarily adjacent) which mark location. Some example from Yucatec:

Code: Select all

Presentative
He’ … =a’       Here it is
He’ … =o’       There it is
He’l … -be’     There it comes (audible), Right here/there

Adverbial
Te’ … =a’        Here
Te’ … =o’        There
Ti’ … =i’          There (anaphoric)
Way … =e’      (In) here
Tol … =o’        (Out) there

Determiner / adnominal
Le … =a’        This
Le … =o’        That
Le … =e’        As for that one, topical 
(Note that le seems to be more or less a definite article, so any common unpossessed definite NP triggers one of these clause final particles)

Some examples:

He’l hun-p’iit ts’aak=a’!
Presentative one-bit cure=deictic1
“Here’s some medicine!”

Kaa=h-ook le=x-chuup chak u=nook=o’
kaa=PRV-enter(B3SG) DET=F-female red(B3SG) 3=garment=deictic2
“And then the woman dressed in red entered…”

It also seems to be true, although it’s hard to find a detailed description of this phenomenon, that these clause final clitics don’t stack, so if multiple triggers for a deictic particle are present in the clause then only one will control the deictic clausal enclitic. E.g. in the following there is only one enclitic, not two:

Ts’a’ le=ba’l te’l=o’!
Give/put(B3SG) DET=thing there=deictic2

Both the determiner le and the adverbial te’/te’l are triggers for clause final deictics, but only one =o’ surfaces. The domain for this one deictic rule includes the pre-verbal focus, if applicable, and all post-verbal material, but excludes a dislocated/fronted topic NP, which takes its own deictic enclitic. Some languages have a special enclitic mostly used to mark fronted topics, like =e' in Yucatec.

Some sources, e.g. Routledge’s The Mayan Languages, briefly describe this as a kind of circumfixal NP definite marking, but this is clearly wrong from descriptions elsewhere (e.g. Demonstratives in Cross-Linguistic Perspective) which make clear that (i) the second element is always placed clause finally even if that doesn’t coincide with the end of the NP, and (ii) elements other definite articles trigger and require the final deictic enclitics.

Anyway, I just thought this was interestingly different enough that others might want to make use of it.

[bradrn]

In some Mayan languages, deictic distinctions (this/that, here/there, …) are not marked by individual words or morphemes but are split between a distance neutral deictic trigger word which determines the function (presentative, adverbial, adnominal, …) and a series of clause final particles (not necessarily adjacent) which mark location. Some example from Yucatec:

Fascinating, thanks for sharing! It reminds me very much of how many languages have fully transparent and compositional sets of pro-forms (see Bhat’s Pronouns), but this is something altogether more weird.

[chris_notts]

In some Mayan languages, deictic distinctions (this/that, here/there, …) are not marked by individual words or morphemes but are split between a distance neutral deictic trigger word which determines the function (presentative, adverbial, adnominal, …) and a series of clause final particles (not necessarily adjacent) which mark location. Some example from Yucatec:

Fascinating, thanks for sharing! It reminds me very much of how many languages have fully transparent and compositional sets of pro-forms (see Bhat’s Pronouns), but this is something altogether more weird.

Yeah, it's definitely fairly unique.

[chris_notts]

I've been thinking about phrase level accents recently. Parts of my current conlang were based on Murrinhpatha, and Jon Mansfield describes an interesting phrase-level accent. He analyses the accent in terms of the PWord and PPhrase, where:

A PWord is a minimally bimoraic morpheme that can host a (high tone) accent on its penultimate syllable. I say can, because when multiple PWords combine into a complex PPhrase, only the rightmost PWord is accented. PWords can be morphologically complex, but not everything written as one unit on paper (without spaces) constitutes a single PWord.

A PPhrase is a morphosyntactically defined unit which contains one or more PWords and zero or more non-PWord adjuncts (clitics and/or affixes that cannot take the accent, basically). The PPhrase has exactly one accent, on the penultimate syllable of the final PWord in the phrase. In simple cases a PPhrase is made up of a single PWord and whatever stress-excluding affixes/clitics are attached to it, but more complex PPhrases contain multiple PWords NPs of the form NEG NOUN+ ADJ DET.

The second example is similar but with rightward tone spreading. According to Moira Yip, in isolation Huave has a rule of high tone insertion on the final (stressed) syllable of words. But when those words are combined into phrases, it is phrasal heads that are accented, with the high tone spreading rightwards to the end of the phrase. How far this extends depends on the phrase type:

* NPs are right-headed, so they are not really affected by spread and just end in a high tone
* The verb phrase forms a unit for this rule, and since Huave is a VO language, high tone spreads from the final syllable of the verb to cover the entire object NP, if present. (Yip says it's unclear if tone spreads to other argument types since all the examples are of VO spreading)

This made me think about a rule for my current conlang which would be a kind of synthesis or adjustment of the two above examples. In the conlang, the verb is effectively an auxiliary-(incorporands)-main.verb compound which occur in rigid order, but with a certain amount of phonological independence. There are basically two units here, aux vs everything else, since the incorporand-main.verb relationship is tight and derivational. I think something like this is not uncommon in ~polysynthetic languages, since I've seen a number of cases of phonological boundaries within the verbal word mentioned. Anyway, I'm imagining the rules that:

* Phonological words can host an accent on their final syllable
* that the verbal grammatical word (a PPhrase in the Murrinhpatha terminology) has a single surface accent, with the accent from the syntactic head dominating if there are multiple phonological words
* The auxiliary complex is the head, but is only accented if it is polysyllabic
* H tone spreads right from the accented syllable to the end of the verb / PPhrase

Since many of the most common prefix chains / aux forms are monosyllabic (mostly if all of positive polarity, not subordinate, 3->3 or SAP->3, stationary/no directionals, common TAM form), this sets up two possible high tone patterns, one minimal (right edge + clitics only), and one maximal (final syllable of prefix chain + incorporands + verb). E.g.:

sisavíq
ʃ-i-saq<vi>
1-PFV-see<SAP.PL>
"We saw it/them"

ernuʃenábsáráq
er-nu-ʃ-en-ab-saq<rV>
3-SUB-1-DIR-FUT-see<3.PL>
"Them being about to see me" (a kind of circumstantial future meaning)

NPs in conlang, on the other hand, have word order (PREP) (QUANT) ART NOUN (ART ADJ)*, although with a tendency towards adj-noun compounding. If I assume that role markers and quantifiers are normally unaccented, that means that the head noun gets a high tone that then spreads over any adjectives. Not sure what to do about possessors and relative clauses, which follow and are complex syntactic structures in their own right... just nuking them feels wrong, so maybe tone spreading doesn't go that far?

There might also be some clause level effects here. Clause final position / pre-intonation break favours devoicing and low tone, at least in statements, so maybe H is excluded from utterance final position?

Another possible rule would be an OCP-like rule that, if two underlying accents / Hs are next to each other, the first deletes. That would have the effect of deleting an accent on the final syllable of the verbal auxiliary, if the incorporand-verb compound that follow is monosyllabic. E.g.:

erʃí-sáq (two underlying accents) -> erʃi-sáq (one deleted) -> ersisáq
NOT ersísáq (if first accent were retained)

Although a combination of the OCP rule and the accent polysyllabic aux only rules would mean that the accent falls on the final syllable of the main verb and final clitics if and only if either the aux is monosyllabic or the main verb is, with the trigger for a larger high tone domain being polysyllabic aux + a morphologically complex verb (either plural inflected, or with applicative prefix, or with converb/noun incorporation).

Anyway, just thought phrasal accents + tone spreading were an interesting combination.

[bradrn]

Not sure what to do about possessors and relative clauses, which follow and are complex syntactic structures in their own right... just nuking them feels wrong, so maybe tone spreading doesn't go that far?

I tend to agree that tone shouldn’t spread onto relative clauses. We know in any case that they tend to be somewhat separable from the rest of the NP (the term is ‘right-dislocation’, I think?), and in terms of intonation they also tend to be somewhat separate too. Not so sure about possessors, though I tend to feel they should participate in spreading.

There might also be some clause level effects here. Clause final position / pre-intonation break favours devoicing and low tone, at least in statements, so maybe H is excluded from utterance final position?

It can happen. On the other hand, there’s plenty of languages which push H to the right edge of the word (with Yip having the details as usual).

Although a combination of the OCP rule and the accent polysyllabic aux only rules would mean that the accent falls on the final syllable of the main verb and final clitics if and only if either the aux is monosyllabic or the main verb is, with the trigger for a larger high tone domain being polysyllabic aux + a morphologically complex verb (either plural inflected, or with applicative prefix, or with converb/noun incorporation).

I don’t see how this follows… could you give an example please?

[chris_notts]

I don’t see how this follows… could you give an example please?

I gave this one. There are two underlying accents since the aux is polysyllabic and the verb has one, but since the verb is monosyllabic the two are adjacent. If there were a rule to delete the first of two underlying accents then a monosyllabic verb with no incorporands would always delete a preceding aux accent.

erʃí-sáq (two underlying accents) -> erʃi-sáq (one deleted) -> ersisáq
NOT ersísáq (if first accent were retained)

[bradrn]

I don’t see how this follows… could you give an example please?

I gave this one. There are two underlying accents since the aux is polysyllabic and the verb has one, but since the verb is monosyllabic the two are adjacent. If there were a rule to delete the first of two underlying accents then a monosyllabic verb with no incorporands would always delete a preceding aux accent.

erʃí-sáq (two underlying accents) -> erʃi-sáq (one deleted) -> ersisáq
NOT ersísáq (if first accent were retained)

OK, that makes sense. Didn’t notice that example.