Chris' scratchpad (was: Ch'ubmin)

[chris_notts]

Starting a thread for grammar excerpts etc. related to my current project.

Phonology chapter, previously posted elsewhere here. The phonemes and phonotactics are broadly based on Mayan, although there's some things inspired by other languages in there.

[chris_notts]

The next chapter that I've worked over until it's in a vaguely comprehensible format is the verb morphology chapter here.

And some tables with a fairly exhaustive list of inflection prefix combinations here.

A lot of Mayan languages are quite prefix heavy and it started out quite similar to K'iche' in structure before I started tweaking, but inspiration-wise there's also a few other things in the mix:

Athabaskan - also prefix heavy, morphemes with similar meanings distributed over the different slots, verb divided into conjunct and disjunct zone with phonological differences between them. If I remember correctly:

Coptic - also has variable ordering of TAM marker and agreement marker depending on the TAM category.

Agreement - borrowings from Sierra Popoluca (number agreement in SP really works that way, although they're suffixes not infixes in that case), and originally Hixkaryana although a lot of that got washed out over time.

Algonquian / Murrinhpatha - initial+final / light verb+coverb structure as dominant predicate structure

Yélî Dnye - basic motion verb / deictic distinctions inflection

Subordinate verb forms - partly inspired by West Greenlandic, although the morphological structure is somewhat different.

Ability and Involuntary Action marking - based on analysis of Salish and other languages with a similar distinction between neutral and culminating perfectives like Tagalog. See e.g.:

https://semantics.uchicago.edu/kennedy/ ... ingh98.pdf
https://semantics.uchicago.edu/kennedy/ ... arel05.pdf

That's quite a lot of stuff, but I think I've managed to iron out the kinks and ditch enough redundant features to the point where it doesn't feel too kitchen-sinky or non-orthogonal, just quite synthetic.

I'm actually quite pleased with this one so far. I generally struggle more with morphology than syntax, which is why a lot of the conlangs I stick with either tend to be more syntax heavy or to favour particles and clitics over gluing things together on the page. But I think this feels reasonably cohesive, not too cartesian (morphemes in one slot rule out morphemes in other slots in a sensible way), and gets rid of awkward phonological sequences in a fairly nice looking way.

[chris_notts]

There are a few things I'm still unsure about. One is purpose vs lest clauses, both of which are morphologically marked. At the moment, both can also be negated... but a negative purpose clause is very similar to a lest clause, and a negative lest clause is similar to a purpose clause.

English does show negation with both structures, and I think there is a semantic difference between e.g. doing something to cause something else not to happen (state of affairs as goal) vs doing something to avoid something else happening (state of affairs as default outcome to avoid).

But it's subtle enough that I'm not sure if it would have been worn away in many more morphological systems, in which case maybe I should just prohibit negative lest clauses.

[chris_notts]

I've been thinking about the interaction between aspect and aktionsart and trying to make sure it makes sense. I think the core of it is that the perfective in Ch'ubmin is an elsewhere aspect: the imperfective is used for events which have an ongoing dynamic aspect at the reference time, otherwise the perfective is used (with the restriction, in positive polarity clauses, that the event at least started in the real/hypothetical world being discussed). This means that the perfective covers:

1. States (not dynamic)
2. Completed achievements (no temporal structure, so if they started then they ended)
3. Accomplishments which started and which stopped (but are only implied, not asserted, to have finished)
4. Activities which ended

Which I guess is maybe fairly typical in many respects for a perfective created as a side effect of an imperfective affix developing, if I remember The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World correctly.

I'm not sure if the imperfective should be allowed with states at all, except maybe with a temporary / maintained state reading, since if energy input is required then the state looks more like an activity. It seems like progressive = temporary state with state verbs is at least a marginal thing in some Salishan languages?

The burden of distinguishing and converting between there even types is spread over various other parts of the grammar:

1. Verb root selection: the closed verb root class are mostly stative or activity/manner with telicity and specific effects added by coverbs, although there are some achievement roots
2. Some stative verbs (mostly positional with classificatory and copula uses) can be converted into inchoatives/accomplishments by addition of directional / associated motion inflection and/or reflexive/middle voice marking.
3, Accomplishments and activities can be converted into something like achievements (forms with lexical aspect which assert the culmination itself) via the AIA suffix. Not sure if it makes sense for stative verbs too... I think probably the AIA suffix only attaches to verbs which already describe some kind of dynamic action, so possibly it cannot be used alone to convert a stative to a dynamic verb, but must combine with some other marker of dynamicity?
4. Inherently dynamic verbs can express resultant states mostly as an implication or via adverbs and clitics like "already", or via subordinate (relative tense) forms combined with a main clause
5. Incorporation / non-specific patients can convert telic predicates into atelic ones.

The language is quite poor in prepositions so I am not sure if oblique role marker choice like the for/in test would make sense. I suspect the form would be the same, i.e. there would be a neutral way of expressing the duration of an action, whether it culminated or not.

[bradrn]

Which I guess is maybe fairly typical in many respects for a perfective created as a side effect of an imperfective affix developing, if I remember The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World correctly.

I agree with this assessment — compare English simple vs progressive, which is basically the same system.

[chris_notts]

Which I guess is maybe fairly typical in many respects for a perfective created as a side effect of an imperfective affix developing, if I remember The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World correctly.

I agree with this assessment — compare English simple vs progressive, which is basically the same system.

Yeah, the main differences are that habitual is not associated with the perfective (that's a function of the imperfective) so it's mostly a non-present form except for states, and that it does not require culmination for durative telic events, it just requires that no dynamic activity is ongoing at the reference time, i.e. the reference time must be outside the temporal bounds of the dynamic phase of the event.

[chris_notts]

I'm trying to flesh out the "moods" a bit more, which were broadly based on some kind of mixture of West Greenlandic and languages like Yidiny, amongst others, both of which have inflections which are somewhat structural in nature rather than purely irrealis/subjunctive/... per se.

West Greenlandic

In the case of West Greenlandic, the Routledge Grammar lists:

Primarily main clause

Indicative
Interrogative
Imperative
Optative

Primarily subordinate clause

Causative
Conditional
Contemporative
Participial

The main clause moods primarily distinguish speech acts, with the imperative and optative being more or less different parts of the same paradigm (imperative forms are 3rd person only, optatives are for all persons except 3rd).

In terms of the subordinate forms, the primary functions are:

Causative

• Reason
• Cause
• When
• Consecutive action (i.e. clause chains)
• Sometimes used as a simple past tense in main clauses

Conditional

• Protasis of conditionals, normally in the future
• When

Contemporative

• Simultaneous action
• Manner
• Purpose (with additional affix)
• Concessive
• When

Participial

• Complement clauses

Yidiny

Dixon's grammar, which is not that easy to extract usage details from, lists:

Main clause

Present-future
Past
Imperative

Subordinate clause

Purposive
Dative subordinate
Causal subordinate
Lest

In terms of the subordinate forms, the primary functions are:

Purposive

• In main clause: optative, hortative, obligation, necessity
• Subsequent action
• Purpose
• Result

Dative Subordinate

• Simultaneous action
• Complement

Causal subordinate

• Prior action
• Cause

Lest

• Negative potential consequence

They can also be used as complementation strategies, although most of the examples involved either the dative or purposive. Basically they seem to be relative tense forms for prior, simultaneous and subsequent action, with the expected logical-causal relations as implications and possible readings. The purposive seems to have the irrealis/modal and causal-logical relation most strongly as part of its meaning, rather than as an implicature.

Ch’ubmin

So the system of Ch’ubmin has:

Main clause:

Indicative (Past Perfective / Future / Imperfective)

Primarily dependent:

Consecutive (Perfective clause chaining form)
Subordinate~Participle (Prior / simultaneous / marginal posterior form)
Purpose
Lest
Conditional

Basically, the subordinate forms express complements and various forms of background adverbial clause, and maybe also relative clauses. Purpose clauses are mostly self explanatory but also serve as an imperative / hortative / optative and as complements in some cases. Lest clauses mark potential negative relative future outcomes of not acting. Conditional clauses mark protases.

This is a bit maximal compared to the previous systems. If I try to standardise the labels a bit, the others have the following approximate subordinate clause types:

GREENLANDIC: Prior~chaining, simultaneous~purpose, conditional, complement
YIDINY: Prior, simultaneous, posterior~purpose, lest

There are at least two interesting differences in the case of Ch’ubmin:

Unlike Greenlandic, it has separate forms for clause chaining (consecutive) and for backgrounded adverbials (the participle~subordinate forms)
Unlike Yidiny, it distinguishes purpose and other forms of posterior (clause chaining, participle)

The differences around the consecutive are related to the fact that Ch’ubmin is a VO language where the “medial” verbs follow the independent verb, whereas Greenlandic is an OV language with medial verbs typically preceding. Lots of people have observed that clause chaining is more common in OV languages and clause chains tend to be longer… and I guess it’s partly because in head-final languages, the independent verb at the end aligns with the culmination of the sentence.

If sentences tend to end on a high, then I think that having a form in an OV language which is ambiguous between a backgrounded prior adverbial clause (when, after, since) and a medial clause is more reasonable than having a form in a VO language which is ambiguous between a posterior, maybe backgrounded and maybe irrealis adverbial clause (purpose etc.) and a medial form. If a head-initial language has a medial or narrative form, it feels more necessary to split it from other kinds of adverbial. And it’s precisely in the posterior forms that there’s a lot of splits:



Of the subordinate forms, there are two primarily anterior ones (conditional and participle), one simultaneous one (participle), and four posterior (participle, consecutive, lest, purpose). The posterior ones differ in terms of background (participles are primarily off-timeline background information, the others are asserted), and reality status (consecutive forms are realis, purpose/lest are irrealis).

If I wanted to simplify this system of posterior forms further, I could:

• Not have a consecutive form (clause chaining is rarer in VO languages anyway)
• Not distinguish purpose (foregrounded, posterior, irrealis) from the somewhat marginal future participle/subordinate form (backgrounded, "about to")
• Not have a lest form

I'm not inclined to do the first, since the main reason to have a consecutive in a very prefix heavy language was to get a bit of variety in what the beginning of verbs looks like. I also quite like having lest, although there is also a frustrative~counterfactual clitic and purpose + frustrative feels like it could convey a lest type meaning.

The second is certainly possible... there is a dedicated affix for purpose, whereas the future participle is mostly a side effect of nu- combining with any indicative tense/aspect combination. I.e. one is monomorphemic and the other is just the compositional outcome of multiple morphemes. I'm not very attached to the future participle, it just feels like a cell that "should" be filled given the morphemes exist and aren't competing for the same slots. I get the impression that the Latin future participle could function both as an "about to" form and to express purpose.

Anyway, just musing on whether the current system feels economical enough or not.

[chris_notts]

I'm thinking about just leaving it for now. The four posterior forms aren't that crowded semantically. I think it's probably fine:

CONSECUTIVE

..., veyoya'
ve-yo-i-a'
CONS-VEN-1-COP
And (then) I came

NU- SUBORDINATE FUTURE / "PARTICIPLE"

nuyošaba'
nu-yo-š-ab-a'
SUBR-VEN-1-FUT-COP
Me being about to come, ...
That I was going to come, ...
Since I was going to come, ...

PURPOSE

... yošoya'
VEN-1-PURP-COP
... so/in order that I come

LEST

uqayoya'
uqa-yo-i-a'
LEST-VEN-1-COP
... lest I come

[chris_notts]

I'm a bit uncomfortable with how close the negative imperfective~future is to the positive imperfective though. They're sa'- and ta- respectively with some minor allomorphy. Negative should maybe be slightly more marked than that. Maybe I'll tweak that first instead.

[chris_notts]

Overview of the negative issue. Originally there were the following vaguely negative morphemes:

oːχ - negative past perfective
si - negative imperfective + irrealis + ...
uqa - lest / so that not

I then added sa' as a special case for a merged negative imperfective/future, because otherwise, thanks to sibilant harmony with the stem, it was possible for si- (negative imperfective = I'm not...) to have the same form as ši- (past perfective = I did...). The idea was that sa' was a contraction of *si-ta- = NEG-IMPERF. But sa' itself looks a lot like ta- which maybe isn't great.

So if I wanted to do something else, what could I do? One fun thing about conlanging is retconning morphology. If I look at the three forms I started from, both the lest form and negative perfective contain a back vowel + uvular sequence. This suggests that maybe uqa is actually *uq-ta- NEG??-IMPERF, and oːχ is actually V-uq- for some V. This is especially plausible because /q/ is typically [qχ] in coda position.

And if that's true, then maybe the correct way to reinforce an ambiguous negative si- is with *uq-. That would give uqsi-, which is not legal in the modern language because it has a coda stop. But that could be resolved by:

Lenition = uχsi-
Deletion or merger = usi, uši, uchi-
Metathesis = usqi-, uski-
Metathesis + lenition/deletion = ujqi-, oːqi-, uːqi-

I'm not sketching out a full history of the phonology and morphology here, but I like things to feel like they could be connected.

What do you guys think?

[chris_notts]

Of course, si-uq- could also just give su' or similar instead of sa'... Need to think about if that fits or not.

[bradrn]

Consecutive (Perfective clause chaining form)
Subordinate~Participle (Prior / simultaneous / marginal posterior form)

I’ve been trying to read through this, and I’m not sure what the difference between these are… in the literature I’ve seen, ‘perfective’ is pretty much synonymous with ‘prior’ when it comes to medial verbs (and similarly with ‘imperfective’ and ‘simultaneous’).

… maybe the correct way to reinforce an ambiguous negative si- is with *uq-. That would give uqsi-, which is not legal in the modern language because it has a coda stop. But that could be resolved by:

Lenition = uχsi-
Deletion or merger = usi, uši, uchi-
Metathesis = usqi-, uski-
Metathesis + lenition/deletion = ujqi-, oːqi-, uːqi-

I'm not sketching out a full history of the phonology and morphology here, but I like things to feel like they could be connected.

What do you guys think?

This looks nice! I think I prefer the form usi- and your later suʼ- most, but that’s just my phonaesthetic opinion.

[chris_notts]

I’ve been trying to read through this, and I’m not sure what the difference between these are… in the literature I’ve seen, ‘perfective’ is pretty much synonymous with ‘prior’ when it comes to medial verbs (and similarly with ‘imperfective’ and ‘simultaneous’).

The difference is that firstly the nu-/na- marked subordinate forms, which I called participles above, are relative tense forms used for complements and backgrounded off timeline information, with typical translations like when, since, after...

The consecutive / medial form, which occurs after the main verb because this is a head initial language, is on the main narrative timeline and advances it.

So the main distinction is: is this clause foregrounded/asserted or backgrounded and is it on the main timeline or off timeline? If it's off timeline, is it before or after the reference time?

If you compare perfective/prior participle and consecutive, then one is a prior and one is a posterior form, and one is naturally backgrounded and off timeline whereas the other is foregrounded and on timeline. It's the difference between:

Prior/perfective particle + main clause
Having eaten/since I had eaten/..., I left.

Main clause + consecutive
I ate and (then) I left.

I guess the first would also fail the "that's not true!" test, i.e. if you say "that's not true!" in response you're almost certainly denying the foregrounded main clause and not that the backgrounded eating event happened.

If you compare imperfective and simultaneous particle then there's a direct analogy in English between an -ing participle and a full progressive clause, or a non-finite complement:

Strolling through the garden, she sang a song
She was strolling through the garden and she sang a song

She likes singing songs
*She likes she sings songs / *she likes she is singing songs

[chris_notts]

There's a quote by Foley on the distinction between medial clauses and subordinate background clauses in The Papuan Languages of New Guinea that's quite good, but also quite long. I'm not sure whether I can type it up or take a picture of the relevant 3-4 pages...

[chris_notts]

Here's some relevant sections:

All Papuan languages, either of the chaining type or otherwise, have subordinate clauses. In the Iatmul example at the beginning of the previous section, there is a form y@-nt-@y-an come-M-UNREAL-if "if he comes", which I termed a subordinate-dependent verb, abbreviated as 'subordinate verb'.

These features of Iatmul are more or less general for Papuan languages. Most Papuan languages have subordinate verb forms which are similar to independent verbs, in that they may be independently inflected for tense/status, a feature we saw in the previous section to be denied to dependent verbs.

For illocutionary force, the greatest differences in behaviour are found. Neither dependent nor subordinate verbs may be inflected for illocutionary force, but this disguises an important difference. Dependent verbs take their illocutionary force specification from the independent verb: whatever speech act is marked on the independent verb applies to the dependent verb as well.

fu-mo k-mi-ga-ka da-pe?
pig-TP 2SG.U-give-DA.OTHER-2SG.A eat.2SG-Q.2SG
"did he give you the pork and did you eat it?"

Subordinate verbs are fundamentally different. While they too may never be inflected for illocutionary force, they are impervious to the illocutionary force specification of the independent verb, always remaining assertions:

fu-mo k-mi-ma-ka da-pe?
pig-TP 2SG.U-give-SUB.OTHER-2SG.A eat.2SG-Q.2SG
"He gave you pork, and so did you eat it?"

This difference in behaviour under the illocutionary force operator provides the fundamental clue to the function of subordinate clauses, as against dependent clauses. The illocutionary force of a subordinate clause may never be anything but a statement, but dependent clauses are not so restricted. Essentially, subordinate clauses represent give, presupposed information (Haiman 1978), a background to the ongoing discourse, while dependent clauses, like independent ones, are part of the ongoing development of discourse, the focused and asserted progression of events. The independent selection of tense/aspect in subordinate ones, but not dependent ones, becomes explicable in light of this difference. Subordinate clauses, in providing background information out of the main line of discourse, need not be tied to the time of discourse, and, further, in order to be sufficiently flexible in expressing kinds of background information, need to be able to select their own tense/status specifications. The time of a background event to a discourse need not be, and often is not, the same as the main time of the discourse. Dependent clauses, on the other hand, constitute the main, asserted lines of the discourse, all of which must occur in the same time frame as the independent verb, on which tense is marked.

Subordinate clauses in Papuan languages always function to background given information, and correspond to two different constructions in more familar languages, adverbial clauses and relative clauses. Both of these constructions express background information in a sentence: adverbial clauses provide a temporal setting or background condition or cause for the asserted new information in the main clause, e.g. while I sat under a tree, an apple hit me on the head or if he comes, I will leave; while relative clauses provide crucial background information for the identification of the referents of their head nouns: the pig which has white spots or the pig which I killed. The close relationship of these two types of subordinate clauses is readily apparent in many Papuan languages, in which they are formally similar or even identical.

[chris_notts]

Not all languages with something like clause chaining seem to firmly split main timeline and backgrounded information. The West Greenlandic system I mentioned earlier doesn't, since the same prior inflection expresses backgrounded information (when, cause) and sequenced events on the main timeline. But the idea with the Ch'ubmin was to approximately split them, partly because if medial verbs and subordinate verbs were confused in a VO language with independent verb first, you'd be confusing real events asserted to happen with irrealis purpose / "going to" forms, since all are posterior. Going back to this table:



The subordinate or participle forms, broadly speaking, show the same tense/aspect distinctions as indicative main clauses, but with time references interpreted relative to the main clause. Their function is adverbial clauses, complement clauses, and maybe / I'm still debating, as the verb forms in relative clauses. They are normally off timeline (in the sense that they don't advance that timeline even if the imperfective forms are simultaneous with it) and backgrounded. Similarly, the conditional is a backgrounded subordinate form dedicated to expressing protases.

The consecutive, which follows the main verb (VO language) and is therefore a posterior form, does advance the main timeline and expresses foregrounded events.

Purpose and lest are foregrounded/asserted (this seems to be true cross-linguistically) but they are not on the main timeline in the sense that the events they describe are not asserted to be realised even if the clause they're subordinate to is realis. They don't advance the timeline or match the mood / reality status of their parent clause in the same way that a consecutive verb has to.

[chris_notts]

This looks nice! I think I prefer the form usi- and your later suʼ- most, but that’s just my phonaesthetic opinion.

I decided to go with su'- because keeping the same CV' shape fit best with the surrounding morphology. It does mean that all ~negative morphemes contain a back vowel (3/4) and/or s (2/4) and/or a uvular (2/4) and most of the options contain multiple of those things, which makes things feel like they cohere a bit:

Code: Select all

oːχ    2     (origin = ?? + uq)
si      1     (origin = si)
uqa   2     (origin = uq + ta)
suˀ    2     (origin = si + uq)

[chris_notts]

I decided to switch to the term "conjunct" for a relative-adverbial-complement verb form, since it seems to be used that way in some Amerindian languages, and it might be less confusing than other terms I was trying like relative, participle, ...

Picking the right terminology to avoid confusing people in any way is hard.

[bradrn]

Picking the right terminology to avoid confusing people in any way is hard.

In linguistics, it’s impossible. Sometimes I give up and pick the most confusing term instead.

[chris_notts]

Picking the right terminology to avoid confusing people in any way is hard.

In linguistics, it’s impossible. Sometimes I give up and pick the most confusing term instead.

So things like "negative: that's what I call the distant past tense because it's for events not happening anymore"